[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16991910#comment-16991910
]
Jonathan Hung commented on HDFS-14973:
--
Removing 2.11.0 fix version after branch-2 -> branch-2.10 rename
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Fix For: 3.3.0, 3.1.4, 3.2.2, 2.10.1
>
> Attachments: HDFS-14973-branch-2.003.patch,
> HDFS-14973-branch-2.004.patch, HDFS-14973-branch-2.005.patch,
> HDFS-14973.000.patch, HDFS-14973.001.patch, HDFS-14973.002.patch,
> HDFS-14973.003.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16983042#comment-16983042
]
Erik Krogen commented on HDFS-14973:
Just committed v005 to branch-2 (and branch-2.10 for good measure, since the
discussion on the mailing list hasn't finalized yet AFAICT). Thanks for the
reviews [~shv]!
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Fix For: 3.3.0, 3.1.4, 3.2.2, 2.10.1, 2.11.0
>
> Attachments: HDFS-14973-branch-2.003.patch,
> HDFS-14973-branch-2.004.patch, HDFS-14973-branch-2.005.patch,
> HDFS-14973.000.patch, HDFS-14973.001.patch, HDFS-14973.002.patch,
> HDFS-14973.003.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16983030#comment-16983030
]
Konstantin Shvachko commented on HDFS-14973:
+1 for v05 patch.
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Fix For: 3.3.0, 3.1.4, 3.2.2
>
> Attachments: HDFS-14973-branch-2.003.patch,
> HDFS-14973-branch-2.004.patch, HDFS-14973-branch-2.005.patch,
> HDFS-14973.000.patch, HDFS-14973.001.patch, HDFS-14973.002.patch,
> HDFS-14973.003.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16982827#comment-16982827
]
Hadoop QA commented on HDFS-14973:
--
| (x) *{color:red}-1 overall{color}* |
\\
\\
|| Vote || Subsystem || Runtime || Comment ||
| {color:blue}0{color} | {color:blue} reexec {color} | {color:blue} 17m
5s{color} | {color:blue} Docker mode activated. {color} |
|| || || || {color:brown} Prechecks {color} ||
| {color:green}+1{color} | {color:green} @author {color} | {color:green} 0m
0s{color} | {color:green} The patch does not contain any @author tags. {color} |
| {color:green}+1{color} | {color:green} test4tests {color} | {color:green} 0m
0s{color} | {color:green} The patch appears to include 3 new or modified test
files. {color} |
|| || || || {color:brown} branch-2 Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 14m
44s{color} | {color:green} branch-2 passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 1m
2s{color} | {color:green} branch-2 passed with JDK v1.7.0_95 {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 1m
9s{color} | {color:green} branch-2 passed with JDK v1.8.0_222 {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
52s{color} | {color:green} branch-2 passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 1m
7s{color} | {color:green} branch-2 passed {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 2m
21s{color} | {color:green} branch-2 passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 1m
20s{color} | {color:green} branch-2 passed with JDK v1.7.0_95 {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
52s{color} | {color:green} branch-2 passed with JDK v1.8.0_222 {color} |
|| || || || {color:brown} Patch Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 1m
7s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 1m
0s{color} | {color:green} the patch passed with JDK v1.7.0_95 {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 1m
0s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
55s{color} | {color:green} the patch passed with JDK v1.8.0_222 {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 0m
55s{color} | {color:green} the patch passed {color} |
| {color:orange}-0{color} | {color:orange} checkstyle {color} | {color:orange}
0m 41s{color} | {color:orange} hadoop-hdfs-project/hadoop-hdfs: The patch
generated 1 new + 644 unchanged - 1 fixed = 645 total (was 645) {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 1m
7s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} whitespace {color} | {color:green} 0m
0s{color} | {color:green} The patch has no whitespace issues. {color} |
| {color:green}+1{color} | {color:green} xml {color} | {color:green} 0m
1s{color} | {color:green} The patch has no ill-formed XML file. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 2m
41s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 1m
19s{color} | {color:green} the patch passed with JDK v1.7.0_95 {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
48s{color} | {color:green} the patch passed with JDK v1.8.0_222 {color} |
|| || || || {color:brown} Other Tests {color} ||
| {color:red}-1{color} | {color:red} unit {color} | {color:red} 86m 9s{color}
| {color:red} hadoop-hdfs in the patch failed. {color} |
| {color:green}+1{color} | {color:green} asflicense {color} | {color:green} 0m
35s{color} | {color:green} The patch does not generate ASF License warnings.
{color} |
| {color:black}{color} | {color:black} {color} | {color:black}139m 41s{color} |
{color:black} {color} |
\\
\\
|| Reason || Tests ||
| Failed junit tests | hadoop.hdfs.server.namenode.TestDeleteRace |
| | hadoop.hdfs.server.datanode.fsdataset.impl.TestLazyPersistReplicaRecovery
|
| | hadoop.hdfs.server.namenode.ha.TestSeveralNameNodes |
| | hadoop.hdfs.qjournal.server.TestJournalNodeRespectsBindHostKeys |
\\
\\
|| Subsystem || Report/Notes ||
| Docker | Client=19.03.5 Server=19.03.5 Image:yetus/hadoop:f555aa740b5 |
| JIRA Issue | HDFS-14973 |
| JIRA Patch URL |
https://issues.apache.org/jira/secure/attachment/12986827/HDFS-14973-branch-2.005.patch
|
| Optional Tests | dupname asflicense compile javac javadoc mvninst
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16982721#comment-16982721
]
Erik Krogen commented on HDFS-14973:
Thanks for the review [~shv]! I addressed your comments in
[^HDFS-14973-branch-2.005.patch] :
# Great idea, thanks.
# I updated it to avoid the use of {{Preconditions}}. It still throws an
{{IllegalArgumentException}}, as I think this is a much better semantic
representation of the issue than an {{IOException}}. The description of an IAE:
"Thrown to indicate that a method has been passed an illegal or inappropriate
argument."
# Yup, totally makes sense. My mistake. Fixed.
# I had a bug in the implementation of {{setAtomicLongToMinMax()}} that would
cause the {{FSNamesystem}} spy to hang sometimes. I fixed this.
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Fix For: 3.3.0, 3.1.4, 3.2.2
>
> Attachments: HDFS-14973-branch-2.003.patch,
> HDFS-14973-branch-2.004.patch, HDFS-14973-branch-2.005.patch,
> HDFS-14973.000.patch, HDFS-14973.001.patch, HDFS-14973.002.patch,
> HDFS-14973.003.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16982062#comment-16982062
]
Konstantin Shvachko commented on HDFS-14973:
Erik, the implementation of {{RateLimiter}} looks good. Few suggestions.
# {{RateLimiter}} looks like a pretty generic tool. We may want to place it
under {{org.apache.hadoop.util}}.
# I would prefer not to use {{Preconditions}}. If it is an error let's throw
an explicit {{IOException}}. If it should not happen, then an assert is
appropriate.
# Should we make {{RateLimiter}} API more similar to guava {{RateLimiter}}, to
minimize version diversion?
{code:java}
public double acquire() // rather than void
public static RateLimiter create(double permitsPerSecond) // rather than int
as parameter
{code}
# I ran {{TestBalancerRPCDelay}} on my linux box, it hangs at the end. Don't
know why.
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Fix For: 3.3.0, 3.1.4, 3.2.2
>
> Attachments: HDFS-14973-branch-2.003.patch,
> HDFS-14973-branch-2.004.patch, HDFS-14973.000.patch, HDFS-14973.001.patch,
> HDFS-14973.002.patch, HDFS-14973.003.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16976939#comment-16976939
]
Hadoop QA commented on HDFS-14973:
--
| (x) *{color:red}-1 overall{color}* |
\\
\\
|| Vote || Subsystem || Runtime || Comment ||
| {color:blue}0{color} | {color:blue} reexec {color} | {color:blue} 1m
21s{color} | {color:blue} Docker mode activated. {color} |
|| || || || {color:brown} Prechecks {color} ||
| {color:green}+1{color} | {color:green} @author {color} | {color:green} 0m
0s{color} | {color:green} The patch does not contain any @author tags. {color} |
| {color:green}+1{color} | {color:green} test4tests {color} | {color:green} 0m
0s{color} | {color:green} The patch appears to include 3 new or modified test
files. {color} |
|| || || || {color:brown} branch-2 Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 10m
33s{color} | {color:green} branch-2 passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 1m
4s{color} | {color:green} branch-2 passed with JDK v1.7.0_95 {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 1m
1s{color} | {color:green} branch-2 passed with JDK v1.8.0_222 {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
46s{color} | {color:green} branch-2 passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 1m
9s{color} | {color:green} branch-2 passed {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 2m
25s{color} | {color:green} branch-2 passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 1m
25s{color} | {color:green} branch-2 passed with JDK v1.7.0_95 {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
57s{color} | {color:green} branch-2 passed with JDK v1.8.0_222 {color} |
|| || || || {color:brown} Patch Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 1m
2s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
59s{color} | {color:green} the patch passed with JDK v1.7.0_95 {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 0m
59s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
56s{color} | {color:green} the patch passed with JDK v1.8.0_222 {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 0m
56s{color} | {color:green} the patch passed {color} |
| {color:orange}-0{color} | {color:orange} checkstyle {color} | {color:orange}
0m 39s{color} | {color:orange} hadoop-hdfs-project/hadoop-hdfs: The patch
generated 1 new + 644 unchanged - 1 fixed = 645 total (was 645) {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 1m
5s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} whitespace {color} | {color:green} 0m
0s{color} | {color:green} The patch has no whitespace issues. {color} |
| {color:green}+1{color} | {color:green} xml {color} | {color:green} 0m
2s{color} | {color:green} The patch has no ill-formed XML file. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 2m
36s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 1m
15s{color} | {color:green} the patch passed with JDK v1.7.0_95 {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
51s{color} | {color:green} the patch passed with JDK v1.8.0_222 {color} |
|| || || || {color:brown} Other Tests {color} ||
| {color:red}-1{color} | {color:red} unit {color} | {color:red} 80m 21s{color}
| {color:red} hadoop-hdfs in the patch failed. {color} |
| {color:green}+1{color} | {color:green} asflicense {color} | {color:green} 0m
26s{color} | {color:green} The patch does not generate ASF License warnings.
{color} |
| {color:black}{color} | {color:black} {color} | {color:black}113m 40s{color} |
{color:black} {color} |
\\
\\
|| Reason || Tests ||
| Failed junit tests |
hadoop.hdfs.qjournal.server.TestJournalNodeRespectsBindHostKeys |
| | hadoop.hdfs.TestRollingUpgrade |
\\
\\
|| Subsystem || Report/Notes ||
| Docker | Client=19.03.5 Server=19.03.5 Image:yetus/hadoop:f555aa740b5 |
| JIRA Issue | HDFS-14973 |
| JIRA Patch URL |
https://issues.apache.org/jira/secure/attachment/12986154/HDFS-14973-branch-2.004.patch
|
| Optional Tests | dupname asflicense compile javac javadoc mvninstall
mvnsite unit shadedclient findbugs checkstyle xml |
| uname | Linux 5837485df68e 4.15.0-66-generic #75-Ubuntu SMP Tue Oct 1
05:24:09 UTC 2019
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16976854#comment-16976854
]
Erik Krogen commented on HDFS-14973:
Attach v004 patch for branch-2 addressing the checkstyle and license issues.
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Fix For: 3.3.0, 3.1.4, 3.2.2
>
> Attachments: HDFS-14973-branch-2.003.patch,
> HDFS-14973-branch-2.004.patch, HDFS-14973.000.patch, HDFS-14973.001.patch,
> HDFS-14973.002.patch, HDFS-14973.003.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16976826#comment-16976826
]
Hadoop QA commented on HDFS-14973:
--
| (x) *{color:red}-1 overall{color}* |
\\
\\
|| Vote || Subsystem || Runtime || Comment ||
| {color:blue}0{color} | {color:blue} reexec {color} | {color:blue} 0m
51s{color} | {color:blue} Docker mode activated. {color} |
|| || || || {color:brown} Prechecks {color} ||
| {color:green}+1{color} | {color:green} @author {color} | {color:green} 0m
0s{color} | {color:green} The patch does not contain any @author tags. {color} |
| {color:green}+1{color} | {color:green} test4tests {color} | {color:green} 0m
0s{color} | {color:green} The patch appears to include 3 new or modified test
files. {color} |
|| || || || {color:brown} branch-2 Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 9m
4s{color} | {color:green} branch-2 passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
54s{color} | {color:green} branch-2 passed with JDK v1.7.0_95 {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
46s{color} | {color:green} branch-2 passed with JDK v1.8.0_222 {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
38s{color} | {color:green} branch-2 passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
55s{color} | {color:green} branch-2 passed {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 2m
4s{color} | {color:green} branch-2 passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 1m
10s{color} | {color:green} branch-2 passed with JDK v1.7.0_95 {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
43s{color} | {color:green} branch-2 passed with JDK v1.8.0_222 {color} |
|| || || || {color:brown} Patch Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 0m
49s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
50s{color} | {color:green} the patch passed with JDK v1.7.0_95 {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 0m
50s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
43s{color} | {color:green} the patch passed with JDK v1.8.0_222 {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 0m
43s{color} | {color:green} the patch passed {color} |
| {color:orange}-0{color} | {color:orange} checkstyle {color} | {color:orange}
0m 34s{color} | {color:orange} hadoop-hdfs-project/hadoop-hdfs: The patch
generated 2 new + 644 unchanged - 1 fixed = 646 total (was 645) {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
54s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} whitespace {color} | {color:green} 0m
0s{color} | {color:green} The patch has no whitespace issues. {color} |
| {color:green}+1{color} | {color:green} xml {color} | {color:green} 0m
2s{color} | {color:green} The patch has no ill-formed XML file. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 2m
6s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 1m
10s{color} | {color:green} the patch passed with JDK v1.7.0_95 {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 0m
47s{color} | {color:green} the patch passed with JDK v1.8.0_222 {color} |
|| || || || {color:brown} Other Tests {color} ||
| {color:red}-1{color} | {color:red} unit {color} | {color:red} 82m 16s{color}
| {color:red} hadoop-hdfs in the patch failed. {color} |
| {color:red}-1{color} | {color:red} asflicense {color} | {color:red} 0m
28s{color} | {color:red} The patch generated 2 ASF License warnings. {color} |
| {color:black}{color} | {color:black} {color} | {color:black}110m 11s{color} |
{color:black} {color} |
\\
\\
|| Reason || Tests ||
| Failed junit tests |
hadoop.hdfs.server.namenode.snapshot.TestSnapshotFileLength |
| | hadoop.hdfs.qjournal.server.TestJournalNodeRespectsBindHostKeys |
\\
\\
|| Subsystem || Report/Notes ||
| Docker | Client=19.03.5 Server=19.03.5 Image:yetus/hadoop:f555aa740b5 |
| JIRA Issue | HDFS-14973 |
| JIRA Patch URL |
https://issues.apache.org/jira/secure/attachment/12986141/HDFS-14973-branch-2.003.patch
|
| Optional Tests | dupname asflicense compile javac javadoc mvninstall
mvnsite unit shadedclient findbugs checkstyle xml |
| uname | Linux 3502fbd10de3 4.15.0-66-generic #75-Ubuntu SMP Tue Oct 1
05:
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16976729#comment-16976729
]
Erik Krogen commented on HDFS-14973:
Just attached a branch-2 patch. There are two changes from the 3.x line:
* Removed the use of {{AtomicLong#getAndUpdate}} within {{TestBalancer}} in
favor of the more primitive {{compareAndSet}} to be Java 7 compatible.
* Remove the use of Guava's {{RateLimiter}}, not present in the version of
Guava used in branch-2. Implemented a simple {{RateLimiter}} that disperses
calls by ensuring they are separated by least (1/rate) time. This can slightly
over-limit. For example with a 10 QPS rate limit, in the following scenario,
thread B will be forced to wait 100ms. I don't think it's an issue in this
case, as all of the dispatch threads will be issued simultaneously, so after a
single initial pause they will all attempt to submit their RPC and be nicely
smoothed out.
** 500ms passes with no RPCs
** thread A submits an RPC
** thread B immediately tries to submit an RPC
[~shv], let me know what you think.
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Fix For: 3.3.0, 3.1.4, 3.2.2
>
> Attachments: HDFS-14973-branch-2.003.patch, HDFS-14973.000.patch,
> HDFS-14973.001.patch, HDFS-14973.002.patch, HDFS-14973.003.patch,
> HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16975301#comment-16975301
]
Hudson commented on HDFS-14973:
---
SUCCESS: Integrated in Jenkins build Hadoop-trunk-Commit #17645 (See
[https://builds.apache.org/job/Hadoop-trunk-Commit/17645/])
HDFS-14973. More strictly enforce Balancer/Mover/SPS throttling of (xkrogen:
rev b2cc8b6b4a78f31cdd937dc4d1a2255f80c5881e)
* (edit)
hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/balancer/TestBalancerRPCDelay.java
* (edit)
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/balancer/Dispatcher.java
* (edit)
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/DFSConfigKeys.java
* (edit) hadoop-hdfs-project/hadoop-hdfs/src/main/resources/hdfs-default.xml
* (edit)
hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/balancer/TestBalancer.java
* (edit)
hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/balancer/NameNodeConnector.java
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Fix For: 3.3.0, 3.2.2
>
> Attachments: HDFS-14973.000.patch, HDFS-14973.001.patch,
> HDFS-14973.002.patch, HDFS-14973.003.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16975299#comment-16975299
]
Erik Krogen commented on HDFS-14973:
Thanks for the review [~shv]. Just committed to trunk and branch-3.2, working
on other backports.
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Fix For: 3.3.0, 3.2.2
>
> Attachments: HDFS-14973.000.patch, HDFS-14973.001.patch,
> HDFS-14973.002.patch, HDFS-14973.003.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16974828#comment-16974828
]
Hadoop QA commented on HDFS-14973:
--
| (x) *{color:red}-1 overall{color}* |
\\
\\
|| Vote || Subsystem || Runtime || Comment ||
| {color:blue}0{color} | {color:blue} reexec {color} | {color:blue} 0m
41s{color} | {color:blue} Docker mode activated. {color} |
|| || || || {color:brown} Prechecks {color} ||
| {color:green}+1{color} | {color:green} @author {color} | {color:green} 0m
0s{color} | {color:green} The patch does not contain any @author tags. {color} |
| {color:green}+1{color} | {color:green} test4tests {color} | {color:green} 0m
0s{color} | {color:green} The patch appears to include 2 new or modified test
files. {color} |
|| || || || {color:brown} trunk Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 19m
11s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 1m
1s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
52s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 1m
5s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
14m 34s{color} | {color:green} branch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 2m
15s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 1m
13s{color} | {color:green} trunk passed {color} |
|| || || || {color:brown} Patch Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 1m
0s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
55s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 0m
55s{color} | {color:green} the patch passed {color} |
| {color:orange}-0{color} | {color:orange} checkstyle {color} | {color:orange}
0m 48s{color} | {color:orange} hadoop-hdfs-project/hadoop-hdfs: The patch
generated 1 new + 699 unchanged - 1 fixed = 700 total (was 700) {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
59s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} whitespace {color} | {color:green} 0m
0s{color} | {color:green} The patch has no whitespace issues. {color} |
| {color:green}+1{color} | {color:green} xml {color} | {color:green} 0m
1s{color} | {color:green} The patch has no ill-formed XML file. {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
13m 24s{color} | {color:green} patch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 2m
21s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 1m
11s{color} | {color:green} the patch passed {color} |
|| || || || {color:brown} Other Tests {color} ||
| {color:red}-1{color} | {color:red} unit {color} | {color:red} 99m 20s{color}
| {color:red} hadoop-hdfs in the patch failed. {color} |
| {color:green}+1{color} | {color:green} asflicense {color} | {color:green} 0m
33s{color} | {color:green} The patch does not generate ASF License warnings.
{color} |
| {color:black}{color} | {color:black} {color} | {color:black}161m 12s{color} |
{color:black} {color} |
\\
\\
|| Reason || Tests ||
| Failed junit tests |
hadoop.hdfs.tools.offlineEditsViewer.TestOfflineEditsViewer |
| | hadoop.hdfs.server.aliasmap.TestSecureAliasMap |
| | hadoop.hdfs.TestDFSUpgradeFromImage |
\\
\\
|| Subsystem || Report/Notes ||
| Docker | Client=19.03.5 Server=19.03.5 Image:yetus/hadoop:104ccca9169 |
| JIRA Issue | HDFS-14973 |
| JIRA Patch URL |
https://issues.apache.org/jira/secure/attachment/12985872/HDFS-14973.003.patch |
| Optional Tests | dupname asflicense compile javac javadoc mvninstall
mvnsite unit shadedclient findbugs checkstyle xml |
| uname | Linux c4b92acf86d2 4.15.0-66-generic #75-Ubuntu SMP Tue Oct 1
05:24:09 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux |
| Build tool | maven |
| Personality | /testptch/patchprocess/precommit/personality/provided.sh |
| git revision | trunk / d0302d3 |
| maven | version: Apache Maven 3.3.9 |
| Default Java | 1.8.0_222 |
| findbugs | v3.1.0-RC1 |
| checkstyle |
https://builds.apache.org/job/PreCommit-HDFS-Build/28312/artifact/out/diff-checkstyle-hadoop-hdfs-project_hadoop-hdfs.txt
|
| uni
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16974759#comment-16974759
]
Konstantin Shvachko commented on HDFS-14973:
This is the Jenkins build that built the patch, but couldn't send the report to
hits jira for some reason:
https://builds.apache.org/job/PreCommit-HDFS-Build/28311/testReport/
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Attachments: HDFS-14973.000.patch, HDFS-14973.001.patch,
> HDFS-14973.002.patch, HDFS-14973.003.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16974756#comment-16974756
]
Konstantin Shvachko commented on HDFS-14973:
+1 Good to go.
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Attachments: HDFS-14973.000.patch, HDFS-14973.001.patch,
> HDFS-14973.002.patch, HDFS-14973.003.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16974418#comment-16974418
]
Erik Krogen commented on HDFS-14973:
Thanks for the comments [~shv], all great points. I put up patch v003
addressing them.
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Attachments: HDFS-14973.000.patch, HDFS-14973.001.patch,
> HDFS-14973.002.patch, HDFS-14973.003.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16973892#comment-16973892
]
Konstantin Shvachko commented on HDFS-14973:
(1) Good point [~xkrogen] . I think Balancer, Mover, and SPS should all be
limited to the same rate of getBlocks to NameNode. This is probably another
good reason to track it inside {{NameNodeConnector}}. Would be good to reflect
this in the description in {{hdfs-site.xml}}
(2) Sure let's leave refactoring of TestBalancer for the next time.
Nits:
# {{dispatchBlockMoves()}} does not need {{dSec}} anymore.
# Seems would be good for {{dispatchBlockMoves()}} to still log
{{concurrentThreads}} in debug mode.
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Attachments: HDFS-14973.000.patch, HDFS-14973.001.patch,
> HDFS-14973.002.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16973838#comment-16973838
]
Hadoop QA commented on HDFS-14973:
--
| (x) *{color:red}-1 overall{color}* |
\\
\\
|| Vote || Subsystem || Runtime || Comment ||
| {color:blue}0{color} | {color:blue} reexec {color} | {color:blue} 0m
56s{color} | {color:blue} Docker mode activated. {color} |
|| || || || {color:brown} Prechecks {color} ||
| {color:green}+1{color} | {color:green} @author {color} | {color:green} 0m
0s{color} | {color:green} The patch does not contain any @author tags. {color} |
| {color:green}+1{color} | {color:green} test4tests {color} | {color:green} 0m
0s{color} | {color:green} The patch appears to include 3 new or modified test
files. {color} |
|| || || || {color:brown} trunk Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 20m
18s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 1m
0s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
55s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 1m
5s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
14m 42s{color} | {color:green} branch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 2m
16s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 1m
12s{color} | {color:green} trunk passed {color} |
|| || || || {color:brown} Patch Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 1m
3s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
55s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 0m
55s{color} | {color:green} the patch passed {color} |
| {color:orange}-0{color} | {color:orange} checkstyle {color} | {color:orange}
0m 50s{color} | {color:orange} hadoop-hdfs-project/hadoop-hdfs: The patch
generated 1 new + 825 unchanged - 1 fixed = 826 total (was 826) {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 0m
59s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} whitespace {color} | {color:green} 0m
1s{color} | {color:green} The patch has no whitespace issues. {color} |
| {color:green}+1{color} | {color:green} xml {color} | {color:green} 0m
1s{color} | {color:green} The patch has no ill-formed XML file. {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
13m 23s{color} | {color:green} patch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 2m
19s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 1m
10s{color} | {color:green} the patch passed {color} |
|| || || || {color:brown} Other Tests {color} ||
| {color:red}-1{color} | {color:red} unit {color} | {color:red}100m 55s{color}
| {color:red} hadoop-hdfs in the patch failed. {color} |
| {color:green}+1{color} | {color:green} asflicense {color} | {color:green} 0m
33s{color} | {color:green} The patch does not generate ASF License warnings.
{color} |
| {color:black}{color} | {color:black} {color} | {color:black}164m 15s{color} |
{color:black} {color} |
\\
\\
|| Reason || Tests ||
| Failed junit tests |
hadoop.hdfs.tools.offlineEditsViewer.TestOfflineEditsViewer |
| | hadoop.hdfs.tools.TestDFSZKFailoverController |
\\
\\
|| Subsystem || Report/Notes ||
| Docker | Client=19.03.4 Server=19.03.4 Image:yetus/hadoop:104ccca9169 |
| JIRA Issue | HDFS-14973 |
| JIRA Patch URL |
https://issues.apache.org/jira/secure/attachment/12985792/HDFS-14973.002.patch |
| Optional Tests | dupname asflicense compile javac javadoc mvninstall
mvnsite unit shadedclient findbugs checkstyle xml |
| uname | Linux c687d780da15 4.15.0-66-generic #75-Ubuntu SMP Tue Oct 1
05:24:09 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux |
| Build tool | maven |
| Personality | /testptch/patchprocess/precommit/personality/provided.sh |
| git revision | trunk / 73a386a |
| maven | version: Apache Maven 3.3.9 |
| Default Java | 1.8.0_222 |
| findbugs | v3.1.0-RC1 |
| checkstyle |
https://builds.apache.org/job/PreCommit-HDFS-Build/28309/artifact/out/diff-checkstyle-hadoop-hdfs-project_hadoop-hdfs.txt
|
| unit |
https://builds.apache.org/job/PreCommit-
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16973732#comment-16973732
]
Erik Krogen commented on HDFS-14973:
For (1), if we do this it will force the {{ExternalStoragePolicySatisfier}},
{{Mover}}, and {{Balancer}} all to use the same rate limiting (as they all make
use of the {{NameNodeConnector}}. I'm not sure if this is desirable or not, any
thoughts on that? This is why I passed it in externally.
For (2), I took a look, as I generally am in agreement with you that not
recreating miniclusters is a good thing. I don't see a good way to do it here,
since {{doTest}} makes use of the ability to configure the exact capacities of
the DataNodes it starts up in the new cluster. I tried refactoring
{{TestBalancer}} a bit to make it so that there is a {{setupMiniCluster}} step
followed by a {{doTestWithoutClusterSetup}} step, but then it becomes very
difficult to properly fill the cluster _only_ on the original DataNodes and not
the new DataNode, leaving a single one empty. Let me know if you have really
strong feelings about this and I can try some more, but for now it seems like
it would require some substantial refactoring of {{TestBalanacer}}.
For running against the default parameter of 20, it wasn't possible with the
existing code since it asserts that the number of getBlocks calls is _higher_
than the rate limit. I changed it to greater-than-or-equal-to instead of a
strict greater-than to get it to work with the default of 20. Uploaded v002
patch with this change.
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Attachments: HDFS-14973.000.patch, HDFS-14973.001.patch,
> HDFS-14973.002.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
--
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16971969#comment-16971969
]
Konstantin Shvachko commented on HDFS-14973:
Anyways, using {{RateLimiter}} sounds like a productive idea. That way we
guarantee {{getBlocks}} do not exceed {{max-qps}} on the NameNode at any time.
Some comments:
# {{NameNodeConnector}} already gets {{Configuration}} as a parameter. Should
we use the config to obtain {{max-qps}} inside the constructor, rather than
dragging it through all calls?
# {{TestBalancer}} is a heavy test since it recreates mini-cluster for each
test case. At least for {{TestBalancerRPCDelay}} can we restructure code so
that it would work against the same mini-cluster, without restarting it.
** Also would be good to run the test against the default parameter once: 20
instead of 10?
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Attachments: HDFS-14973.000.patch, HDFS-14973.001.patch,
> HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16971876#comment-16971876
]
Hadoop QA commented on HDFS-14973:
--
| (x) *{color:red}-1 overall{color}* |
\\
\\
|| Vote || Subsystem || Runtime || Comment ||
| {color:blue}0{color} | {color:blue} reexec {color} | {color:blue} 0m
50s{color} | {color:blue} Docker mode activated. {color} |
|| || || || {color:brown} Prechecks {color} ||
| {color:green}+1{color} | {color:green} @author {color} | {color:green} 0m
0s{color} | {color:green} The patch does not contain any @author tags. {color} |
| {color:green}+1{color} | {color:green} test4tests {color} | {color:green} 0m
0s{color} | {color:green} The patch appears to include 3 new or modified test
files. {color} |
|| || || || {color:brown} trunk Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 19m
12s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
59s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
54s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 1m
6s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
15m 6s{color} | {color:green} branch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 2m
14s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 1m
16s{color} | {color:green} trunk passed {color} |
|| || || || {color:brown} Patch Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 1m
4s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
55s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 0m
55s{color} | {color:green} the patch passed {color} |
| {color:orange}-0{color} | {color:orange} checkstyle {color} | {color:orange}
0m 50s{color} | {color:orange} hadoop-hdfs-project/hadoop-hdfs: The patch
generated 1 new + 825 unchanged - 1 fixed = 826 total (was 826) {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 1m
1s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} whitespace {color} | {color:green} 0m
0s{color} | {color:green} The patch has no whitespace issues. {color} |
| {color:green}+1{color} | {color:green} xml {color} | {color:green} 0m
1s{color} | {color:green} The patch has no ill-formed XML file. {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
13m 26s{color} | {color:green} patch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 2m
21s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 1m
8s{color} | {color:green} the patch passed {color} |
|| || || || {color:brown} Other Tests {color} ||
| {color:red}-1{color} | {color:red} unit {color} | {color:red} 98m 45s{color}
| {color:red} hadoop-hdfs in the patch failed. {color} |
| {color:green}+1{color} | {color:green} asflicense {color} | {color:green} 0m
33s{color} | {color:green} The patch does not generate ASF License warnings.
{color} |
| {color:black}{color} | {color:black} {color} | {color:black}161m 27s{color} |
{color:black} {color} |
\\
\\
|| Reason || Tests ||
| Failed junit tests | hadoop.hdfs.TestDFSInotifyEventInputStreamKerberized |
| | hadoop.hdfs.tools.TestDFSZKFailoverController |
\\
\\
|| Subsystem || Report/Notes ||
| Docker | Client=19.03.4 Server=19.03.4 Image:yetus/hadoop:104ccca9169 |
| JIRA Issue | HDFS-14973 |
| JIRA Patch URL |
https://issues.apache.org/jira/secure/attachment/12985545/HDFS-14973.001.patch |
| Optional Tests | dupname asflicense compile javac javadoc mvninstall
mvnsite unit shadedclient findbugs checkstyle xml |
| uname | Linux 9c7ff84a7583 4.15.0-66-generic #75-Ubuntu SMP Tue Oct 1
05:24:09 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux |
| Build tool | maven |
| Personality | /testptch/patchprocess/precommit/personality/provided.sh |
| git revision | trunk / 30b93f9 |
| maven | version: Apache Maven 3.3.9 |
| Default Java | 1.8.0_222 |
| findbugs | v3.1.0-RC1 |
| checkstyle |
https://builds.apache.org/job/PreCommit-HDFS-Build/28288/artifact/out/diff-checkstyle-hadoop-hdfs-project_hadoop-hdfs.txt
|
| unit |
https://builds.apache.org/job/PreCommit-HDFS-Bui
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16971813#comment-16971813
]
Erik Krogen commented on HDFS-14973:
Thanks for taking a look [~shv]!
{quote}The next wave of dispatcher threads after 100 should not hit the
NameNode right away. It is supposed first to executePendingMove(), then call
getBlocks(). And executePendingMove() naturally throttles the dispatcher, so it
was not necessary to delay the subsequent ways.
{quote}
This might work, except that {{executePendingMove}} is non-blocking:
{code:java}
public void executePendingMove(final PendingMove p) {
// move the reportedBlock
final DDatanode targetDn = p.target.getDDatanode();
ExecutorService moveExecutor = targetDn.getMoveExecutor();
if (moveExecutor == null) {
final int nThreads = moverThreadAllocator.allocate();
if (nThreads > 0) {
moveExecutor = targetDn.initMoveExecutor(nThreads);
}
}
if (moveExecutor == null) {
LOG.warn("No mover threads available: skip moving " + p);
targetDn.removePendingBlock(p);
p.proxySource.removePendingBlock(p);
return;
}
moveExecutor.execute(new Runnable() {
@Override
public void run() {
p.dispatch();
}
});
}
{code}
It simply allocates a thread pool (if one does not exist), then submits a task
to it to be executed. The actual movement will be executed later, by the
{{moveExecutor}}. Even as far back as 2.6.1 (which doesn't have HDFS-11742),
{{executePendingMove}} was similarly nonblocking:
{code:java}
public void executePendingMove(final PendingMove p) {
// move the block
moveExecutor.execute(new Runnable() {
@Override
public void run() {
p.dispatch();
}
});
}
{code}
However I certainly agree that it's possible the changes to the balancer
(HDFS-8818, HDFS-11742) exacerbated this issue.
With this non-blocking behavior, you end up with the scenario I described where
the first 20 slots in the {{dispatchExecutor}} continue to push through
dispatch tasks with throughput above the throttling limit.
Attached v1 patch addressing hdfs-site.xml issue.
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Attachments: HDFS-14973.000.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16970673#comment-16970673
]
Konstantin Shvachko commented on HDFS-14973:
Hey [~xkrogen] good analysis.
IIRC, the idea was to delay the first "wave" of getBlocks. The first 100 in
your example, which is the number dispatcher threads. Indeed the first 20 will
go right away without delay. This is how many calls we want to tolerate on the
NameNode at once. One second later another 20 getBlocks() will hit the
NameNode, and so on up to 100.
The next wave of dispatcher threads after 100 should not hit the NameNode right
away. It is supposed first to {{executePendingMove()}}, then call
{{getBlocks()}}. And {{executePendingMove()}} naturally throttles the
dispatcher, so it was not necessary to delay the subsequent ways. I remember it
worked. It is possible that {{executePendingMove()}} became faster due to
HDFS-11742, which I did not check, or something else had changed.
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Attachments: HDFS-14973.000.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> (time to submit getBlocks RPCs) + (DataNode startup time) + (time for the
> Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16970633#comment-16970633
]
Hadoop QA commented on HDFS-14973:
--
| (x) *{color:red}-1 overall{color}* |
\\
\\
|| Vote || Subsystem || Runtime || Comment ||
| {color:blue}0{color} | {color:blue} reexec {color} | {color:blue} 0m
50s{color} | {color:blue} Docker mode activated. {color} |
|| || || || {color:brown} Prechecks {color} ||
| {color:green}+1{color} | {color:green} @author {color} | {color:green} 0m
0s{color} | {color:green} The patch does not contain any @author tags. {color} |
| {color:green}+1{color} | {color:green} test4tests {color} | {color:green} 0m
0s{color} | {color:green} The patch appears to include 3 new or modified test
files. {color} |
|| || || || {color:brown} trunk Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 19m
13s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 1m
2s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} checkstyle {color} | {color:green} 0m
53s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 1m
5s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
14m 36s{color} | {color:green} branch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 2m
17s{color} | {color:green} trunk passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 1m
13s{color} | {color:green} trunk passed {color} |
|| || || || {color:brown} Patch Compile Tests {color} ||
| {color:green}+1{color} | {color:green} mvninstall {color} | {color:green} 1m
0s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} compile {color} | {color:green} 0m
55s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javac {color} | {color:green} 0m
55s{color} | {color:green} the patch passed {color} |
| {color:orange}-0{color} | {color:orange} checkstyle {color} | {color:orange}
0m 50s{color} | {color:orange} hadoop-hdfs-project/hadoop-hdfs: The patch
generated 1 new + 825 unchanged - 1 fixed = 826 total (was 826) {color} |
| {color:green}+1{color} | {color:green} mvnsite {color} | {color:green} 1m
2s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} whitespace {color} | {color:green} 0m
0s{color} | {color:green} The patch has no whitespace issues. {color} |
| {color:green}+1{color} | {color:green} shadedclient {color} | {color:green}
14m 1s{color} | {color:green} patch has no errors when building and testing
our client artifacts. {color} |
| {color:green}+1{color} | {color:green} findbugs {color} | {color:green} 2m
40s{color} | {color:green} the patch passed {color} |
| {color:green}+1{color} | {color:green} javadoc {color} | {color:green} 1m
11s{color} | {color:green} the patch passed {color} |
|| || || || {color:brown} Other Tests {color} ||
| {color:red}-1{color} | {color:red} unit {color} | {color:red}108m 45s{color}
| {color:red} hadoop-hdfs in the patch failed. {color} |
| {color:green}+1{color} | {color:green} asflicense {color} | {color:green} 0m
33s{color} | {color:green} The patch does not generate ASF License warnings.
{color} |
| {color:black}{color} | {color:black} {color} | {color:black}172m 1s{color} |
{color:black} {color} |
\\
\\
|| Reason || Tests ||
| Failed junit tests | hadoop.tools.TestHdfsConfigFields |
| | hadoop.hdfs.TestReconstructStripedFile |
| | hadoop.hdfs.server.namenode.TestFSImage |
\\
\\
|| Subsystem || Report/Notes ||
| Docker | Client=19.03.4 Server=19.03.4 Image:yetus/hadoop:104ccca9169 |
| JIRA Issue | HDFS-14973 |
| JIRA Patch URL |
https://issues.apache.org/jira/secure/attachment/12985394/HDFS-14973.000.patch |
| Optional Tests | dupname asflicense compile javac javadoc mvninstall
mvnsite unit shadedclient findbugs checkstyle |
| uname | Linux a91b493a2d4d 4.15.0-66-generic #75-Ubuntu SMP Tue Oct 1
05:24:09 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux |
| Build tool | maven |
| Personality | /testptch/patchprocess/precommit/personality/provided.sh |
| git revision | trunk / 42fc888 |
| maven | version: Apache Maven 3.3.9 |
| Default Java | 1.8.0_222 |
| findbugs | v3.1.0-RC1 |
| checkstyle |
https://builds.apache.org/job/PreCommit-HDFS-Build/28282/artifact/out/diff-checkstyle-hadoop-hdfs-project_hadoop-hdfs.txt
|
| unit |
https://builds.apache.org/job/PreCommit-HDFS-Build/28282/artifact/out/patch-unit-hadoop-hdfs-project_hadoop-hdfs.txt
|
| Test Results |
https://builds.apache.org/job/PreCommit-H
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16970560#comment-16970560
]
Erik Krogen commented on HDFS-14973:
Attached v000 patch with a fix. Conveniently, Guava has a
[{{RateLimiter}}|https://guava.dev/releases/19.0/api/docs/index.html?com/google/common/util/concurrent/RateLimiter.html]
class which does exactly what we need. The changes were minimal.
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Attachments: HDFS-14973.000.patch, HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> 2 * (time to submit getBlocks RPCs) + (DataNode startup time) + 2 * (time for
> the Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-14973) Balancer getBlocks RPC dispersal does not function properly
[
https://issues.apache.org/jira/browse/HDFS-14973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16970500#comment-16970500
]
Erik Krogen commented on HDFS-14973:
I have attached [^HDFS-14973.test.patch] which fixes the test to demonstrate
that the throttling isn't working as expected:
* Adjust the balancing in the test to be performed over only a single
iteration, so that the Dispatcher's block move time isn't counted
* Adjust the time at which the {{startGetBlocksTime}} is initialized, so that
the DataNode startup time isn't counted
* Make the number of max {{getBlocks}} RPCs configurable to have better control
over the test
I will work on putting a fix together. I think we _might_ be able to fix this
by simply starting the delay at 1 second instead of 0 seconds, but I don't
think it would be very hard to have a more strict throttling mechanism to avoid
this entire class of problem, so I'm going to take a stab at that. If it turns
out to be too complex, I'll try something simple.
> Balancer getBlocks RPC dispersal does not function properly
> ---
>
> Key: HDFS-14973
> URL: https://issues.apache.org/jira/browse/HDFS-14973
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: balancer & mover
>Affects Versions: 2.9.0, 2.7.4, 2.8.2, 3.0.0
>Reporter: Erik Krogen
>Assignee: Erik Krogen
>Priority: Major
> Attachments: HDFS-14973.test.patch
>
>
> In HDFS-11384, a mechanism was added to make the {{getBlocks}} RPC calls
> issued by the Balancer/Mover more dispersed, to alleviate load on the
> NameNode, since {{getBlocks}} can be very expensive and the Balancer should
> not impact normal cluster operation.
> Unfortunately, this functionality does not function as expected, especially
> when the dispatcher thread count is low. The primary issue is that the delay
> is applied only to the first N threads that are submitted to the dispatcher's
> executor, where N is the size of the dispatcher's threadpool, but *not* to
> the first R threads, where R is the number of allowed {{getBlocks}} QPS
> (currently hardcoded to 20). For example, if the threadpool size is 100 (the
> default), threads 0-19 have no delay, 20-99 have increased levels of delay,
> and 100+ have no delay. As I understand it, the intent of the logic was that
> the delay applied to the first 100 threads would force the dispatcher
> executor's threads to all be consumed, thus blocking subsequent (non-delayed)
> threads until the delay period has expired. However, threads 0-19 can finish
> very quickly (their work can often be fulfilled in the time it takes to
> execute a single {{getBlocks}} RPC, on the order of tens of milliseconds),
> thus opening up 20 new slots in the executor, which are then consumed by
> non-delayed threads 100-119, and so on. So, although 80 threads have had a
> delay applied, the non-delay threads rush through in the 20 non-delay slots.
> This problem gets even worse when the dispatcher threadpool size is less than
> the max {{getBlocks}} QPS. For example, if the threadpool size is 10, _no
> threads ever have a delay applied_, and the feature is not enabled at all.
> This problem wasn't surfaced in the original JIRA because the test
> incorrectly measured the period across which {{getBlocks}} RPCs were
> distributed. The variables {{startGetBlocksTime}} and {{endGetBlocksTime}}
> were used to track the time over which the {{getBlocks}} calls were made.
> However, {{startGetBlocksTime}} was initialized at the time of creation of
> the {{FSNameystem}} spy, which is before the mock DataNodes are started. Even
> worse, the Balancer in this test takes 2 iterations to complete balancing the
> cluster, so the time period {{endGetBlocksTime - startGetBlocksTime}}
> actually represents:
> {code}
> 2 * (time to submit getBlocks RPCs) + (DataNode startup time) + 2 * (time for
> the Dispatcher to complete an iteration of moving blocks)
> {code}
> Thus, the RPC QPS reported by the test is much lower than the RPC QPS seen
> during the period of initial block fetching.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
-
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
