[jira] [Commented] (HDFS-11153) RPC Client detect address changed should reconnect immediately
[ https://issues.apache.org/jira/browse/HDFS-11153?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16020587#comment-16020587 ] DENG FEI commented on HDFS-11153: - As default behavior,why found host is changed,but do nothing.it's unreasonable,and we can't release very corrected configuration for every client. [~Kihwal Lee] > RPC Client detect address changed should reconnect immediately > -- > > Key: HDFS-11153 > URL: https://issues.apache.org/jira/browse/HDFS-11153 > Project: Hadoop HDFS > Issue Type: Bug > Components: ipc >Affects Versions: 3.0.0-alpha1 >Reporter: DENG FEI > Attachments: HDFS-1153.001.patch > > > HA mode,the _*"ipc.client.connect.max.retries.on.timeouts"*_ and > _*"ipc.client.connect.max.retries"*_ is set zero,but if met active NN's ip > changed,it will detect the change,but won't reconnect because exceed the max > retry times,after do 15 times failover and then throw connection or standby > exception. > maybe if found the address is changed,should reconnect immediately no matter > the retry times limit. > > log is below: > {noformat} > 2016-11-16 17:00:20,844 (WARN org.apache.hadoop.ipc.Client 510): Address > change detected. Old: *:9000 New: X:9000 > 2016-11-16 17:01:09,893 (WARN org.apache.hadoop.ipc.Client 510): Address > change detected. Old: *::9000 New: X:9000 > 2016-11-16 17:01:09,893 (WARN > org.apache.hadoop.io.retry.RetryInvocationHandler 118): Exception while > invoking class > org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolTranslatorPB.getFileInfo. > Not retrying because failovers (15) exceeded maximum allowed (15) > {noformat} -- This message was sent by Atlassian JIRA (v6.3.15#6346) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-11153) RPC Client detect address changed should reconnect immediately
[ https://issues.apache.org/jira/browse/HDFS-11153?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16020593#comment-16020593 ] KWON BYUNGCHANG commented on HDFS-11153: This problem also occurs in ResourceManager. I think this issue seems to fit the hadoop common jira. > RPC Client detect address changed should reconnect immediately > -- > > Key: HDFS-11153 > URL: https://issues.apache.org/jira/browse/HDFS-11153 > Project: Hadoop HDFS > Issue Type: Bug > Components: ipc >Affects Versions: 3.0.0-alpha1 >Reporter: DENG FEI > Attachments: HDFS-1153.001.patch, stupid.png > > > HA mode,the _*"ipc.client.connect.max.retries.on.timeouts"*_ and > _*"ipc.client.connect.max.retries"*_ is set zero,but if met active NN's ip > changed,it will detect the change,but won't reconnect because exceed the max > retry times,after do 15 times failover and then throw connection or standby > exception. > maybe if found the address is changed,should reconnect immediately no matter > the retry times limit. > > log is below: > {noformat} > 2016-11-16 17:00:20,844 (WARN org.apache.hadoop.ipc.Client 510): Address > change detected. Old: *:9000 New: X:9000 > 2016-11-16 17:01:09,893 (WARN org.apache.hadoop.ipc.Client 510): Address > change detected. Old: *::9000 New: X:9000 > 2016-11-16 17:01:09,893 (WARN > org.apache.hadoop.io.retry.RetryInvocationHandler 118): Exception while > invoking class > org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolTranslatorPB.getFileInfo. > Not retrying because failovers (15) exceeded maximum allowed (15) > {noformat} -- This message was sent by Atlassian JIRA (v6.3.15#6346) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-11153) RPC Client detect address changed should reconnect immediately
[ https://issues.apache.org/jira/browse/HDFS-11153?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16057160#comment-16057160 ] DENG FEI commented on HDFS-11153: - at least,should config dfs.client.failover.connection.retries.on.timeouts=1 as default by not 0 > RPC Client detect address changed should reconnect immediately > -- > > Key: HDFS-11153 > URL: https://issues.apache.org/jira/browse/HDFS-11153 > Project: Hadoop HDFS > Issue Type: Bug > Components: ipc >Affects Versions: 3.0.0-alpha1 >Reporter: DENG FEI > Attachments: HDFS-1153.001.patch, stupid.png > > > HA mode,the _*"ipc.client.connect.max.retries.on.timeouts"*_ and > _*"ipc.client.connect.max.retries"*_ is set zero,but if met active NN's ip > changed,it will detect the change,but won't reconnect because exceed the max > retry times,after do 15 times failover and then throw connection or standby > exception. > maybe if found the address is changed,should reconnect immediately no matter > the retry times limit. > > log is below: > {noformat} > 2016-11-16 17:00:20,844 (WARN org.apache.hadoop.ipc.Client 510): Address > change detected. Old: *:9000 New: X:9000 > 2016-11-16 17:01:09,893 (WARN org.apache.hadoop.ipc.Client 510): Address > change detected. Old: *::9000 New: X:9000 > 2016-11-16 17:01:09,893 (WARN > org.apache.hadoop.io.retry.RetryInvocationHandler 118): Exception while > invoking class > org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolTranslatorPB.getFileInfo. > Not retrying because failovers (15) exceeded maximum allowed (15) > {noformat} -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-11153) RPC Client detect address changed should reconnect immediately
[ https://issues.apache.org/jira/browse/HDFS-11153?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15673846#comment-15673846 ] Kihwal Lee commented on HDFS-11153: --- You can enable this through config. Here is an example: {noformat} dfs.client.retry.policy.enabled=true dfs.client.retry.policy.spec=1000,2 {noformat} > RPC Client detect address changed should reconnect immediately > -- > > Key: HDFS-11153 > URL: https://issues.apache.org/jira/browse/HDFS-11153 > Project: Hadoop HDFS > Issue Type: Bug > Components: ipc >Affects Versions: 3.0.0-alpha1 >Reporter: DENG FEI > > HA mode,the _*"ipc.client.connect.max.retries.on.timeouts"*_ and > _*"ipc.client.connect.max.retries"*_ is set zero,but if met active NN's ip > changed,it will detect the change,but won't reconnect because exceed the max > retry times,after do 15 times failover and then throw connection or standby > exception. > maybe if found the address is changed,should reconnect immediately no matter > the retry times limit. > > log is below: > {noformat} > 2016-11-16 17:00:20,844 (WARN org.apache.hadoop.ipc.Client 510): Address > change detected. Old: *:9000 New: X:9000 > 2016-11-16 17:01:09,893 (WARN org.apache.hadoop.ipc.Client 510): Address > change detected. Old: *::9000 New: X:9000 > 2016-11-16 17:01:09,893 (WARN > org.apache.hadoop.io.retry.RetryInvocationHandler 118): Exception while > invoking class > org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolTranslatorPB.getFileInfo. > Not retrying because failovers (15) exceeded maximum allowed (15) > {noformat} -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
[jira] [Commented] (HDFS-11153) RPC Client detect address changed should reconnect immediately
[ https://issues.apache.org/jira/browse/HDFS-11153?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15675579#comment-15675579 ] DENG FEI commented on HDFS-11153: - [~kihwal] In the case ,it's should reconnect irrespective of retry policy maybe better. And the *_ConnectTimeoutException_* not handle by _MultipleLinearRandomRetry_ retry policy. > RPC Client detect address changed should reconnect immediately > -- > > Key: HDFS-11153 > URL: https://issues.apache.org/jira/browse/HDFS-11153 > Project: Hadoop HDFS > Issue Type: Bug > Components: ipc >Affects Versions: 3.0.0-alpha1 >Reporter: DENG FEI > > HA mode,the _*"ipc.client.connect.max.retries.on.timeouts"*_ and > _*"ipc.client.connect.max.retries"*_ is set zero,but if met active NN's ip > changed,it will detect the change,but won't reconnect because exceed the max > retry times,after do 15 times failover and then throw connection or standby > exception. > maybe if found the address is changed,should reconnect immediately no matter > the retry times limit. > > log is below: > {noformat} > 2016-11-16 17:00:20,844 (WARN org.apache.hadoop.ipc.Client 510): Address > change detected. Old: *:9000 New: X:9000 > 2016-11-16 17:01:09,893 (WARN org.apache.hadoop.ipc.Client 510): Address > change detected. Old: *::9000 New: X:9000 > 2016-11-16 17:01:09,893 (WARN > org.apache.hadoop.io.retry.RetryInvocationHandler 118): Exception while > invoking class > org.apache.hadoop.hdfs.protocolPB.ClientNamenodeProtocolTranslatorPB.getFileInfo. > Not retrying because failovers (15) exceeded maximum allowed (15) > {noformat} -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org