Monitoring move progress
I've got some nodes in a moving state in a cluster (the nodes to which they stream shouldn't overlap), and I'm finding it difficult to determine if they're actually doing anything related to the move at this point, or if they're stuck in the state and not actually doing anything. In each case, I issued the move command per usual. The log shows information about the move when it begins, showing the correct token change that I would expect in each case. Compactions took place on each moving node, which can be viewed through nodetool compactionstats or through the CompactionManager in JMX. But eventually the compactions stopped, apart from various ongoing secondary index rebuilds and consequent related index compactions. Yet I see no stream transfers via netstats. My expectation is that after the compactions (which the project wiki refers to as anti-compactions), I would start to see outbound streaming activity in netstats. Yet I do not. I don't see any errors listed in the logs on the moving servers since the moves began. Using cassandra 1.0.5. ByteOrderedPartitioner. Any suggestions on how to determine what's going on? Thanks in advance. - Ethan
New node unable to stream (0.8.5)
Hi. We've been running a 7-node cluster with RF 3, QUORUM reads/writes in our production environment for a few months. It's been consistently stable during this period, particularly once we got out maintenance strategy fully worked out (per node, one repair a week, one major compaction a week, the latter due to the nature of our data model and usage). While this cluster started, back in June or so, on the 0.7 series, it's been running 0.8.3 for a while now with no issues. We upgraded to 0.8.5 two days ago, having tested the upgrade in our staging cluster (with an otherwise identical configuration) previously and verified that our application's various use cases appeared successful. One of our nodes suffered a disk failure yesterday. We attempted to replace the dead node by placing a new node at OldNode.initial_token - 1 with auto_bootstrap on. A few things went awry from there: 1. We never saw the new node in bootstrap mode; it became available pretty much immediately upon joining the ring, and never reported a joining state. I did verify that auto_bootstrap was on. 2. I mistakenly ran repair on the new node rather than removetoken on the old node, due to a delightful mental error. The repair got nowhere fast, as it attempts to repair against the down node which throws an exception. So I interrupted the repair, restarted the node to clear any pending validation compactions, and... 3. Ran removetoken for the old node. 4. We let this run for some time and saw eventually that all the nodes appeared to be done various compactions and were stuck at streaming. Many streams listed as open, none making any progress. 5. I observed an Rpc-related exception on the new node (where the removetoken was launched) and concluded that the streams were broken so the process wouldn't ever finish. 6. Ran a removetoken force to get the dead node out of the mix. No problems. 7. Ran a repair on the new node. 8. Validations ran, streams opened up, and again things got stuck in streaming, hanging for over an hour with no progress. 9. Musing that lingering tasks from the removetoken could be a factor, I performed a rolling restart and attempted a repair again. 10. Same problem. Did another rolling restart and attempted a fresh repair on the most important column family alone. 11. Same problem. Streams included CFs not specified, so I guess they must be for hinted handoff. In concluding that streaming is stuck, I've observed: - streams will be open to the new node from other nodes, but the new node doesn't list them - streams will be open to the other nodes from the new node, but the other nodes don't list them - the streams reported may make some initial progress, but then they hang at a particular point and do not move on for an hour or more. - The logs report repair-related activity, until NPEs on incoming TCP connections show up, which appear likely to be the culprit. I can provide more exact details when I'm done commuting. With streaming broken on this node, I'm unable to run repairs, which is obviously problematic. The application didn't suffer any operational issues as a consequence of this, but I need to review the overnight results to verify we're not suffering data loss (I doubt we are). At this point, I'm considering a couple options: 1. Remove the new node and let the adjacent node take over its range 2. Bring the new node down, add a new one in front of it, and properly removetoken the problematic one. 3. Bring the new node down, remove all its data except for the system keyspace, then bring it back up and repair it. 4. Revert to 0.8.3 and see if that helps. Recommendations? Thanks. - Ethan
Re: New node unable to stream (0.8.5)
Here's a typical log slice (not terribly informative, I fear): INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,106 AntiEntropyService.java (l ine 884) Performing streaming repair of 1003 ranges with /10.34.90.8 for (299 90798416657667504332586989223299634,54296681768153272037430773234349600451] INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,427 StreamOut.java (line 181) Stream context metadata [/mnt/cassandra/data/events_production/FitsByShip-g-1 0-Data.db sections=88 progress=0/11707163 - 0%, /mnt/cassandra/data/events_pr oduction/FitsByShip-g-11-Data.db sections=169 progress=0/6133240 - 0%, /mnt/c assandra/data/events_production/FitsByShip-g-6-Data.db sections=1 progress=0/ 6918814 - 0%, /mnt/cassandra/data/events_production/FitsByShip-g-12-Data.db s ections=260 progress=0/9091780 - 0%], 4 sstables. INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,428 StreamOutSession.java (lin e 174) Streaming to /10.34.90.8 ERROR [Thread-56] 2011-09-15 05:41:38,515 AbstractCassandraDaemon.java (line 139) Fatal exception in thread Thread[Thread-56,5,main] java.lang.NullPointerException at org.apache.cassandra.net.IncomingTcpConnection.stream(IncomingTcpC onnection.java:174) at org.apache.cassandra.net.IncomingTcpConnection.run(IncomingTcpConn ection.java:114) Not sure if the exception is related to the outbound streaming above; other nodes are actively trying to stream to this node, so perhaps it comes from those and temporal adjacency to the outbound stream is just coincidental. I have other snippets that look basically identical to the above, except if I look at the logs to which this node is trying to stream, I see that it has concurrently opened a stream in the other direction, which could be the one that the exception pertains to. On Thu, Sep 15, 2011 at 7:41 AM, Sylvain Lebresne sylv...@datastax.comwrote: On Thu, Sep 15, 2011 at 1:16 PM, Ethan Rowe et...@the-rowes.com wrote: Hi. We've been running a 7-node cluster with RF 3, QUORUM reads/writes in our production environment for a few months. It's been consistently stable during this period, particularly once we got out maintenance strategy fully worked out (per node, one repair a week, one major compaction a week, the latter due to the nature of our data model and usage). While this cluster started, back in June or so, on the 0.7 series, it's been running 0.8.3 for a while now with no issues. We upgraded to 0.8.5 two days ago, having tested the upgrade in our staging cluster (with an otherwise identical configuration) previously and verified that our application's various use cases appeared successful. One of our nodes suffered a disk failure yesterday. We attempted to replace the dead node by placing a new node at OldNode.initial_token - 1 with auto_bootstrap on. A few things went awry from there: 1. We never saw the new node in bootstrap mode; it became available pretty much immediately upon joining the ring, and never reported a joining state. I did verify that auto_bootstrap was on. 2. I mistakenly ran repair on the new node rather than removetoken on the old node, due to a delightful mental error. The repair got nowhere fast, as it attempts to repair against the down node which throws an exception. So I interrupted the repair, restarted the node to clear any pending validation compactions, and... 3. Ran removetoken for the old node. 4. We let this run for some time and saw eventually that all the nodes appeared to be done various compactions and were stuck at streaming. Many streams listed as open, none making any progress. 5. I observed an Rpc-related exception on the new node (where the removetoken was launched) and concluded that the streams were broken so the process wouldn't ever finish. 6. Ran a removetoken force to get the dead node out of the mix. No problems. 7. Ran a repair on the new node. 8. Validations ran, streams opened up, and again things got stuck in streaming, hanging for over an hour with no progress. 9. Musing that lingering tasks from the removetoken could be a factor, I performed a rolling restart and attempted a repair again. 10. Same problem. Did another rolling restart and attempted a fresh repair on the most important column family alone. 11. Same problem. Streams included CFs not specified, so I guess they must be for hinted handoff. In concluding that streaming is stuck, I've observed: - streams will be open to the new node from other nodes, but the new node doesn't list them - streams will be open to the other nodes from the new node, but the other nodes don't list them - the streams reported may make some initial progress, but then they hang at a particular point and do not move on for an hour or more. - The logs report repair-related activity, until NPEs on incoming TCP connections show up, which appear likely
Re: New node unable to stream (0.8.5)
I just noticed the following from one of Jonathan Ellis' messages yesterday: Added to NEWS: - After upgrading, run nodetool scrub against each node before running repair, moving nodes, or adding new ones. We did not do this, as it was not indicated as necessary in the news when we were dealing with the upgrade. So perhaps I need to scrub everything before going any further, though the question is what to do with the problematic node. Additionally, it would be helpful to know if scrub will affect the hinted handoffs that have accumulated, as these seem likely to be part of the set of failing streams. On Thu, Sep 15, 2011 at 8:13 AM, Ethan Rowe et...@the-rowes.com wrote: Here's a typical log slice (not terribly informative, I fear): INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,106 AntiEntropyService.java (l ine 884) Performing streaming repair of 1003 ranges with /10.34.90.8 for (299 90798416657667504332586989223299634,54296681768153272037430773234349600451] INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,427 StreamOut.java (line 181) Stream context metadata [/mnt/cassandra/data/events_production/FitsByShip-g-1 0-Data.db sections=88 progress=0/11707163 - 0%, /mnt/cassandra/data/events_pr oduction/FitsByShip-g-11-Data.db sections=169 progress=0/6133240 - 0%, /mnt/c assandra/data/events_production/FitsByShip-g-6-Data.db sections=1 progress=0/ 6918814 - 0%, /mnt/cassandra/data/events_production/FitsByShip-g-12-Data.db s ections=260 progress=0/9091780 - 0%], 4 sstables. INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,428 StreamOutSession.java (lin e 174) Streaming to /10.34.90.8 ERROR [Thread-56] 2011-09-15 05:41:38,515 AbstractCassandraDaemon.java (line 139) Fatal exception in thread Thread[Thread-56,5,main] java.lang.NullPointerException at org.apache.cassandra.net.IncomingTcpConnection.stream(IncomingTcpC onnection.java:174) at org.apache.cassandra.net.IncomingTcpConnection.run(IncomingTcpConn ection.java:114) Not sure if the exception is related to the outbound streaming above; other nodes are actively trying to stream to this node, so perhaps it comes from those and temporal adjacency to the outbound stream is just coincidental. I have other snippets that look basically identical to the above, except if I look at the logs to which this node is trying to stream, I see that it has concurrently opened a stream in the other direction, which could be the one that the exception pertains to. On Thu, Sep 15, 2011 at 7:41 AM, Sylvain Lebresne sylv...@datastax.comwrote: On Thu, Sep 15, 2011 at 1:16 PM, Ethan Rowe et...@the-rowes.com wrote: Hi. We've been running a 7-node cluster with RF 3, QUORUM reads/writes in our production environment for a few months. It's been consistently stable during this period, particularly once we got out maintenance strategy fully worked out (per node, one repair a week, one major compaction a week, the latter due to the nature of our data model and usage). While this cluster started, back in June or so, on the 0.7 series, it's been running 0.8.3 for a while now with no issues. We upgraded to 0.8.5 two days ago, having tested the upgrade in our staging cluster (with an otherwise identical configuration) previously and verified that our application's various use cases appeared successful. One of our nodes suffered a disk failure yesterday. We attempted to replace the dead node by placing a new node at OldNode.initial_token - 1 with auto_bootstrap on. A few things went awry from there: 1. We never saw the new node in bootstrap mode; it became available pretty much immediately upon joining the ring, and never reported a joining state. I did verify that auto_bootstrap was on. 2. I mistakenly ran repair on the new node rather than removetoken on the old node, due to a delightful mental error. The repair got nowhere fast, as it attempts to repair against the down node which throws an exception. So I interrupted the repair, restarted the node to clear any pending validation compactions, and... 3. Ran removetoken for the old node. 4. We let this run for some time and saw eventually that all the nodes appeared to be done various compactions and were stuck at streaming. Many streams listed as open, none making any progress. 5. I observed an Rpc-related exception on the new node (where the removetoken was launched) and concluded that the streams were broken so the process wouldn't ever finish. 6. Ran a removetoken force to get the dead node out of the mix. No problems. 7. Ran a repair on the new node. 8. Validations ran, streams opened up, and again things got stuck in streaming, hanging for over an hour with no progress. 9. Musing that lingering tasks from the removetoken could be a factor, I performed a rolling restart and attempted a repair again. 10. Same problem. Did another rolling
Re: New node unable to stream (0.8.5)
After further review, I'm definitely going to scrub all the original nodes in the cluster. We've lost some data as a result of this situation. It can be restored, but the question is what to do with the problematic new node first. I don't particularly care about the data that's on it, since I'm going to re-import the critical data from files anyway, and then I can recreate derivative data afterwards. So it's purely a matter of getting the cluster healthy again as quickly as possible so I can begin that import process. Any issue with running scrubs on multiple nodes at a time, provided they aren't replication neighbors? On Thu, Sep 15, 2011 at 8:18 AM, Ethan Rowe et...@the-rowes.com wrote: I just noticed the following from one of Jonathan Ellis' messages yesterday: Added to NEWS: - After upgrading, run nodetool scrub against each node before running repair, moving nodes, or adding new ones. We did not do this, as it was not indicated as necessary in the news when we were dealing with the upgrade. So perhaps I need to scrub everything before going any further, though the question is what to do with the problematic node. Additionally, it would be helpful to know if scrub will affect the hinted handoffs that have accumulated, as these seem likely to be part of the set of failing streams. On Thu, Sep 15, 2011 at 8:13 AM, Ethan Rowe et...@the-rowes.com wrote: Here's a typical log slice (not terribly informative, I fear): INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,106 AntiEntropyService.java (l ine 884) Performing streaming repair of 1003 ranges with /10.34.90.8 for (299 90798416657667504332586989223299634,54296681768153272037430773234349600451] INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,427 StreamOut.java (line 181) Stream context metadata [/mnt/cassandra/data/events_production/FitsByShip-g-1 0-Data.db sections=88 progress=0/11707163 - 0%, /mnt/cassandra/data/events_pr oduction/FitsByShip-g-11-Data.db sections=169 progress=0/6133240 - 0%, /mnt/c assandra/data/events_production/FitsByShip-g-6-Data.db sections=1 progress=0/ 6918814 - 0%, /mnt/cassandra/data/events_production/FitsByShip-g-12-Data.db s ections=260 progress=0/9091780 - 0%], 4 sstables. INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,428 StreamOutSession.java (lin e 174) Streaming to /10.34.90.8 ERROR [Thread-56] 2011-09-15 05:41:38,515 AbstractCassandraDaemon.java (line 139) Fatal exception in thread Thread[Thread-56,5,main] java.lang.NullPointerException at org.apache.cassandra.net.IncomingTcpConnection.stream(IncomingTcpC onnection.java:174) at org.apache.cassandra.net.IncomingTcpConnection.run(IncomingTcpConn ection.java:114) Not sure if the exception is related to the outbound streaming above; other nodes are actively trying to stream to this node, so perhaps it comes from those and temporal adjacency to the outbound stream is just coincidental. I have other snippets that look basically identical to the above, except if I look at the logs to which this node is trying to stream, I see that it has concurrently opened a stream in the other direction, which could be the one that the exception pertains to. On Thu, Sep 15, 2011 at 7:41 AM, Sylvain Lebresne sylv...@datastax.comwrote: On Thu, Sep 15, 2011 at 1:16 PM, Ethan Rowe et...@the-rowes.com wrote: Hi. We've been running a 7-node cluster with RF 3, QUORUM reads/writes in our production environment for a few months. It's been consistently stable during this period, particularly once we got out maintenance strategy fully worked out (per node, one repair a week, one major compaction a week, the latter due to the nature of our data model and usage). While this cluster started, back in June or so, on the 0.7 series, it's been running 0.8.3 for a while now with no issues. We upgraded to 0.8.5 two days ago, having tested the upgrade in our staging cluster (with an otherwise identical configuration) previously and verified that our application's various use cases appeared successful. One of our nodes suffered a disk failure yesterday. We attempted to replace the dead node by placing a new node at OldNode.initial_token - 1 with auto_bootstrap on. A few things went awry from there: 1. We never saw the new node in bootstrap mode; it became available pretty much immediately upon joining the ring, and never reported a joining state. I did verify that auto_bootstrap was on. 2. I mistakenly ran repair on the new node rather than removetoken on the old node, due to a delightful mental error. The repair got nowhere fast, as it attempts to repair against the down node which throws an exception. So I interrupted the repair, restarted the node to clear any pending validation compactions, and... 3. Ran removetoken for the old node. 4. We let this run for some time and saw eventually that all the nodes appeared to be done various
Re: New node unable to stream (0.8.5)
Thanks, Jonathan. I'll try the workaround and see if that gets the streams flowing properly. As I mentioned before, we did not run scrub yet. What is the consequence of letting the streams from the hinted handoffs complete if scrub hasn't been run on these nodes? I'm currently running scrub on one node to get a sense of the time frame. Thanks again. - Ethan On Thu, Sep 15, 2011 at 9:09 AM, Jonathan Ellis jbel...@gmail.com wrote: That means we missed a place we needed to special-case for backwards compatibility -- the workaround is, add an empty encryption_options section to cassandra.yaml: encryption_options: internode_encryption: none keystore: conf/.keystore keystore_password: cassandra truststore: conf/.truststore truststore_password: cassandra Created https://issues.apache.org/jira/browse/CASSANDRA-3212 to fix this. On Thu, Sep 15, 2011 at 7:13 AM, Ethan Rowe et...@the-rowes.com wrote: Here's a typical log slice (not terribly informative, I fear): INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,106 AntiEntropyService.java (l ine 884) Performing streaming repair of 1003 ranges with /10.34.90.8for (299 90798416657667504332586989223299634,54296681768153272037430773234349600451] INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,427 StreamOut.java (line 181) Stream context metadata [/mnt/cassandra/data/events_production/FitsByShip-g-1 0-Data.db sections=88 progress=0/11707163 - 0%, /mnt/cassandra/data/events_pr oduction/FitsByShip-g-11-Data.db sections=169 progress=0/6133240 - 0%, /mnt/c assandra/data/events_production/FitsByShip-g-6-Data.db sections=1 progress=0/ 6918814 - 0%, /mnt/cassandra/data/events_production/FitsByShip-g-12-Data.db s ections=260 progress=0/9091780 - 0%], 4 sstables. INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,428 StreamOutSession.java (lin e 174) Streaming to /10.34.90.8 ERROR [Thread-56] 2011-09-15 05:41:38,515 AbstractCassandraDaemon.java (line 139) Fatal exception in thread Thread[Thread-56,5,main] java.lang.NullPointerException at org.apache.cassandra.net.IncomingTcpConnection.stream(IncomingTcpC onnection.java:174) at org.apache.cassandra.net.IncomingTcpConnection.run(IncomingTcpConn ection.java:114) Not sure if the exception is related to the outbound streaming above; other nodes are actively trying to stream to this node, so perhaps it comes from those and temporal adjacency to the outbound stream is just coincidental. I have other snippets that look basically identical to the above, except if I look at the logs to which this node is trying to stream, I see that it has concurrently opened a stream in the other direction, which could be the one that the exception pertains to. On Thu, Sep 15, 2011 at 7:41 AM, Sylvain Lebresne sylv...@datastax.com wrote: On Thu, Sep 15, 2011 at 1:16 PM, Ethan Rowe et...@the-rowes.com wrote: Hi. We've been running a 7-node cluster with RF 3, QUORUM reads/writes in our production environment for a few months. It's been consistently stable during this period, particularly once we got out maintenance strategy fully worked out (per node, one repair a week, one major compaction a week, the latter due to the nature of our data model and usage). While this cluster started, back in June or so, on the 0.7 series, it's been running 0.8.3 for a while now with no issues. We upgraded to 0.8.5 two days ago, having tested the upgrade in our staging cluster (with an otherwise identical configuration) previously and verified that our application's various use cases appeared successful. One of our nodes suffered a disk failure yesterday. We attempted to replace the dead node by placing a new node at OldNode.initial_token - 1 with auto_bootstrap on. A few things went awry from there: 1. We never saw the new node in bootstrap mode; it became available pretty much immediately upon joining the ring, and never reported a joining state. I did verify that auto_bootstrap was on. 2. I mistakenly ran repair on the new node rather than removetoken on the old node, due to a delightful mental error. The repair got nowhere fast, as it attempts to repair against the down node which throws an exception. So I interrupted the repair, restarted the node to clear any pending validation compactions, and... 3. Ran removetoken for the old node. 4. We let this run for some time and saw eventually that all the nodes appeared to be done various compactions and were stuck at streaming. Many streams listed as open, none making any progress. 5. I observed an Rpc-related exception on the new node (where the removetoken was launched) and concluded that the streams were broken so the process wouldn't ever finish. 6. Ran a removetoken force to get the dead node out of the mix
Re: New node unable to stream (0.8.5)
On Thu, Sep 15, 2011 at 9:21 AM, Jonathan Ellis jbel...@gmail.com wrote: Where did the data loss come in? The outcome of the analytical jobs run overnight while some of these repairs were (not) running is consistent with what I would expect if perhaps 20-30% of the source data was missing. Given the strong consistency model we're using, this is surprising to me, since the jobs did not report any read or write failures. I wonder if this is a consequence of the dead node missing and the new node being operational but having received basically none of its hinted handoff streams. Perhaps with streaming fixed the data will reappear, which would be a happy outcome, but if not, I can reimport the critical stuff from files. Scrub is safe to run in parallel. Is it somewhat analogous to a major compaction in terms of I/O impact, with perhaps less greedy use of disk space? On Thu, Sep 15, 2011 at 8:08 AM, Ethan Rowe et...@the-rowes.com wrote: After further review, I'm definitely going to scrub all the original nodes in the cluster. We've lost some data as a result of this situation. It can be restored, but the question is what to do with the problematic new node first. I don't particularly care about the data that's on it, since I'm going to re-import the critical data from files anyway, and then I can recreate derivative data afterwards. So it's purely a matter of getting the cluster healthy again as quickly as possible so I can begin that import process. Any issue with running scrubs on multiple nodes at a time, provided they aren't replication neighbors? On Thu, Sep 15, 2011 at 8:18 AM, Ethan Rowe et...@the-rowes.com wrote: I just noticed the following from one of Jonathan Ellis' messages yesterday: Added to NEWS: - After upgrading, run nodetool scrub against each node before running repair, moving nodes, or adding new ones. We did not do this, as it was not indicated as necessary in the news when we were dealing with the upgrade. So perhaps I need to scrub everything before going any further, though the question is what to do with the problematic node. Additionally, it would be helpful to know if scrub will affect the hinted handoffs that have accumulated, as these seem likely to be part of the set of failing streams. On Thu, Sep 15, 2011 at 8:13 AM, Ethan Rowe et...@the-rowes.com wrote: Here's a typical log slice (not terribly informative, I fear): INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,106 AntiEntropyService.java (l ine 884) Performing streaming repair of 1003 ranges with /10.34.90.8for (299 90798416657667504332586989223299634,54296681768153272037430773234349600451] INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,427 StreamOut.java (line 181) Stream context metadata [/mnt/cassandra/data/events_production/FitsByShip-g-1 0-Data.db sections=88 progress=0/11707163 - 0%, /mnt/cassandra/data/events_pr oduction/FitsByShip-g-11-Data.db sections=169 progress=0/6133240 - 0%, /mnt/c assandra/data/events_production/FitsByShip-g-6-Data.db sections=1 progress=0/ 6918814 - 0%, /mnt/cassandra/data/events_production/FitsByShip-g-12-Data.db s ections=260 progress=0/9091780 - 0%], 4 sstables. INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,428 StreamOutSession.java (lin e 174) Streaming to /10.34.90.8 ERROR [Thread-56] 2011-09-15 05:41:38,515 AbstractCassandraDaemon.java (line 139) Fatal exception in thread Thread[Thread-56,5,main] java.lang.NullPointerException at org.apache.cassandra.net.IncomingTcpConnection.stream(IncomingTcpC onnection.java:174) at org.apache.cassandra.net.IncomingTcpConnection.run(IncomingTcpConn ection.java:114) Not sure if the exception is related to the outbound streaming above; other nodes are actively trying to stream to this node, so perhaps it comes from those and temporal adjacency to the outbound stream is just coincidental. I have other snippets that look basically identical to the above, except if I look at the logs to which this node is trying to stream, I see that it has concurrently opened a stream in the other direction, which could be the one that the exception pertains to. On Thu, Sep 15, 2011 at 7:41 AM, Sylvain Lebresne sylv...@datastax.com wrote: On Thu, Sep 15, 2011 at 1:16 PM, Ethan Rowe et...@the-rowes.com wrote: Hi. We've been running a 7-node cluster with RF 3, QUORUM reads/writes in our production environment for a few months. It's been consistently stable during this period, particularly once we got out maintenance strategy fully worked out (per node, one repair a week, one major compaction a week, the latter due to the nature of our data model and usage). While this cluster started, back in June or so, on the 0.7 series, it's been running 0.8.3 for a while now with no issues. We upgraded to 0.8.5 two days ago
Re: New node unable to stream (0.8.5)
On Thu, Sep 15, 2011 at 10:03 AM, Jonathan Ellis jbel...@gmail.com wrote: If you added the new node as a seed, it would ignore bootstrap mode. And bootstrap / repair *do* use streaming so you'll want to re-run repair post-scrub. (No need to re-bootstrap since you're repairing.) Ah, of course. That's what happened; the chef recipe added the node to its own seed list, which is a problem I thought we'd fixed but apparently not. That definitely explains the bootstrap issue. But no matter, so long as the repairs can eventually run. Scrub is a little less heavyweight than major compaction but same ballpark. It runs sstable-at-a-time so (as long as you haven't been in the habit of forcing majors) space should not be a concern. Cool. We've deactivated all tasks against these nodes and will scrub them all in parallel, apply the encryption options you specified, and see where that gets us. Thanks for the assistance. - Ethan On Thu, Sep 15, 2011 at 8:40 AM, Ethan Rowe et...@the-rowes.com wrote: On Thu, Sep 15, 2011 at 9:21 AM, Jonathan Ellis jbel...@gmail.com wrote: Where did the data loss come in? The outcome of the analytical jobs run overnight while some of these repairs were (not) running is consistent with what I would expect if perhaps 20-30% of the source data was missing. Given the strong consistency model we're using, this is surprising to me, since the jobs did not report any read or write failures. I wonder if this is a consequence of the dead node missing and the new node being operational but having received basically none of its hinted handoff streams. Perhaps with streaming fixed the data will reappear, which would be a happy outcome, but if not, I can reimport the critical stuff from files. Scrub is safe to run in parallel. Is it somewhat analogous to a major compaction in terms of I/O impact, with perhaps less greedy use of disk space? On Thu, Sep 15, 2011 at 8:08 AM, Ethan Rowe et...@the-rowes.com wrote: After further review, I'm definitely going to scrub all the original nodes in the cluster. We've lost some data as a result of this situation. It can be restored, but the question is what to do with the problematic new node first. I don't particularly care about the data that's on it, since I'm going to re-import the critical data from files anyway, and then I can recreate derivative data afterwards. So it's purely a matter of getting the cluster healthy again as quickly as possible so I can begin that import process. Any issue with running scrubs on multiple nodes at a time, provided they aren't replication neighbors? On Thu, Sep 15, 2011 at 8:18 AM, Ethan Rowe et...@the-rowes.com wrote: I just noticed the following from one of Jonathan Ellis' messages yesterday: Added to NEWS: - After upgrading, run nodetool scrub against each node before running repair, moving nodes, or adding new ones. We did not do this, as it was not indicated as necessary in the news when we were dealing with the upgrade. So perhaps I need to scrub everything before going any further, though the question is what to do with the problematic node. Additionally, it would be helpful to know if scrub will affect the hinted handoffs that have accumulated, as these seem likely to be part of the set of failing streams. On Thu, Sep 15, 2011 at 8:13 AM, Ethan Rowe et...@the-rowes.com wrote: Here's a typical log slice (not terribly informative, I fear): INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,106 AntiEntropyService.java (l ine 884) Performing streaming repair of 1003 ranges with / 10.34.90.8 for (299 90798416657667504332586989223299634,54296681768153272037430773234349600451] INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,427 StreamOut.java (line 181) Stream context metadata [/mnt/cassandra/data/events_production/FitsByShip-g-1 0-Data.db sections=88 progress=0/11707163 - 0%, /mnt/cassandra/data/events_pr oduction/FitsByShip-g-11-Data.db sections=169 progress=0/6133240 - 0%, /mnt/c assandra/data/events_production/FitsByShip-g-6-Data.db sections=1 progress=0/ 6918814 - 0%, /mnt/cassandra/data/events_production/FitsByShip-g-12-Data.db s ections=260 progress=0/9091780 - 0%], 4 sstables. INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,428 StreamOutSession.java (lin e 174) Streaming to /10.34.90.8 ERROR [Thread-56] 2011-09-15 05:41:38,515 AbstractCassandraDaemon.java (line 139) Fatal exception in thread Thread[Thread-56,5,main] java.lang.NullPointerException at org.apache.cassandra.net.IncomingTcpConnection.stream(IncomingTcpC onnection.java:174) at org.apache.cassandra.net.IncomingTcpConnection.run(IncomingTcpConn ection.java:114) Not sure if the exception is related to the outbound
Re: New node unable to stream (0.8.5)
Cool. We've deactivated all tasks against these nodes and will scrub them all in parallel, apply the encryption options you specified, and see where that gets us. Thanks for the assistance. To follow up: * We scrubbed all the nodes * We applied the encryption options specified * A repair is continuing (for about an hour so far, perhaps more) on the new, problematic node; it's successfully streaming data from its neighbors and has built up a roughly equivalent data volume on disk We'll see if the data is fully restored once this process completes. Even if it isn't, it seems likely that the cluster will be in a healthy state soon, so we can reimport as necessary and we'll be out of the woods. Now that I've said all that, something will inevitably go wrong, but until that happens, thanks again for the feedback. - Ethan On Thu, Sep 15, 2011 at 8:40 AM, Ethan Rowe et...@the-rowes.com wrote: On Thu, Sep 15, 2011 at 9:21 AM, Jonathan Ellis jbel...@gmail.com wrote: Where did the data loss come in? The outcome of the analytical jobs run overnight while some of these repairs were (not) running is consistent with what I would expect if perhaps 20-30% of the source data was missing. Given the strong consistency model we're using, this is surprising to me, since the jobs did not report any read or write failures. I wonder if this is a consequence of the dead node missing and the new node being operational but having received basically none of its hinted handoff streams. Perhaps with streaming fixed the data will reappear, which would be a happy outcome, but if not, I can reimport the critical stuff from files. Scrub is safe to run in parallel. Is it somewhat analogous to a major compaction in terms of I/O impact, with perhaps less greedy use of disk space? On Thu, Sep 15, 2011 at 8:08 AM, Ethan Rowe et...@the-rowes.com wrote: After further review, I'm definitely going to scrub all the original nodes in the cluster. We've lost some data as a result of this situation. It can be restored, but the question is what to do with the problematic new node first. I don't particularly care about the data that's on it, since I'm going to re-import the critical data from files anyway, and then I can recreate derivative data afterwards. So it's purely a matter of getting the cluster healthy again as quickly as possible so I can begin that import process. Any issue with running scrubs on multiple nodes at a time, provided they aren't replication neighbors? On Thu, Sep 15, 2011 at 8:18 AM, Ethan Rowe et...@the-rowes.com wrote: I just noticed the following from one of Jonathan Ellis' messages yesterday: Added to NEWS: - After upgrading, run nodetool scrub against each node before running repair, moving nodes, or adding new ones. We did not do this, as it was not indicated as necessary in the news when we were dealing with the upgrade. So perhaps I need to scrub everything before going any further, though the question is what to do with the problematic node. Additionally, it would be helpful to know if scrub will affect the hinted handoffs that have accumulated, as these seem likely to be part of the set of failing streams. On Thu, Sep 15, 2011 at 8:13 AM, Ethan Rowe et...@the-rowes.com wrote: Here's a typical log slice (not terribly informative, I fear): INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,106 AntiEntropyService.java (l ine 884) Performing streaming repair of 1003 ranges with / 10.34.90.8 for (299 90798416657667504332586989223299634,54296681768153272037430773234349600451] INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,427 StreamOut.java (line 181) Stream context metadata [/mnt/cassandra/data/events_production/FitsByShip-g-1 0-Data.db sections=88 progress=0/11707163 - 0%, /mnt/cassandra/data/events_pr oduction/FitsByShip-g-11-Data.db sections=169 progress=0/6133240 - 0%, /mnt/c assandra/data/events_production/FitsByShip-g-6-Data.db sections=1 progress=0/ 6918814 - 0%, /mnt/cassandra/data/events_production/FitsByShip-g-12-Data.db s ections=260 progress=0/9091780 - 0%], 4 sstables. INFO [AntiEntropyStage:2] 2011-09-15 05:41:36,428 StreamOutSession.java (lin e 174) Streaming to /10.34.90.8 ERROR [Thread-56] 2011-09-15 05:41:38,515 AbstractCassandraDaemon.java (line 139) Fatal exception in thread Thread[Thread-56,5,main] java.lang.NullPointerException at org.apache.cassandra.net.IncomingTcpConnection.stream(IncomingTcpC onnection.java:174) at org.apache.cassandra.net.IncomingTcpConnection.run(IncomingTcpConn ection.java:114) Not sure if the exception is related to the outbound streaming above; other nodes are actively trying to stream to this node, so perhaps it comes from
Truncate introspection
Is there any straightforward means of seeing what's going on after issuing a truncate (on 0.7.5)? I'm not seeing evidence that anything actually happened. I've disabled read repair on the column family in question and don't have anything actively reading/writing at present, apart from my one-off tests to see if rows have disappeared. Thanks in advance.
Re: Truncate introspection
If those went to zero, it would certainly tell me something happened. :) I guess watching that would be a way of seeing something was going on. Is the truncate itself propagating a ring-wide marker or anything so the CF is logically empty before being physically removed? That's the impression I got from the docs but it wasn't totally clear to me. On Mon, Jun 27, 2011 at 3:33 PM, Jonathan Ellis jbel...@gmail.com wrote: There's a JMX method to get the number of sstables in a CF, is that what you're looking for? On Mon, Jun 27, 2011 at 1:04 PM, Ethan Rowe et...@the-rowes.com wrote: Is there any straightforward means of seeing what's going on after issuing a truncate (on 0.7.5)? I'm not seeing evidence that anything actually happened. I've disabled read repair on the column family in question and don't have anything actively reading/writing at present, apart from my one-off tests to see if rows have disappeared. Thanks in advance. -- Jonathan Ellis Project Chair, Apache Cassandra co-founder of DataStax, the source for professional Cassandra support http://www.datastax.com
Re: What's the best modeling approach for ordering events by date?
Hi. So, the OPP will direct all activity for a range of keys to a particular node (or set of nodes, in accordance with your replication factor). Depending on the volume of writes, this could be fine. Depending on the distribution of key values you write at any given time, it can also be fine. But if you're using the OPP, and your keys align with the time of receiving the data, and your application writes that data as it receives it, you're going to be placing write activity on effectively one node at a time, for the range of time allocated to that node. If you use RP, and can divide time into finer slices such that you have multiple tweets in a row, you trade off a more complex read in exchange for better distribution of load throughout your cluster. The necessity of this depends on your particulars. In your TweetsBySecond example, you're using a deterministic set of keys (the keys correspond to seconds since epoch). Querying for ranges of time is nice with OPP, but if the ranges of time you're interested in are constrained, you don't specifically need OPP. You could use RP and request all the keys for the seconds contained within the time range of interest. In this way, you balance writes across the cluster more effectively than you would with OPP, while still getting a workable data set. Again, the degree to which you need this is dependent on your situation. Others on the list will no doubt have more informed opinions on this than me. :) On Thu, Apr 14, 2011 at 8:00 PM, Guillermo Winkler gwink...@inconcertcc.com wrote: Hi Ethan, I want to present the events ordered by time, always in pages of 20/40 events. If the events are tweets, you can have 1000 tweets from the same second or you can have 30 tweets in a 10 minute range. But I always wanna be able to page through the results in an orderly fashion. I think that using seconds since epoch it's what I'm doing, that is divide time into a fixed series of interval. Each second is an interval, and all of the events for that particular second are columns of that row. Again with tweets for easier visualizatoin TweetsBySecond : { 12121121212 :{ - seconds since epoch id1,id2,id3 - all the tweet ids ocurred in that particular second }, 12121212123 : { id4,id5 }, 12121212124 : { id6 } } The problem is you can't do that using OPP in cassandra 0.7, or it's just me missing something? Thanks for your answer, Guille On Thu, Apr 14, 2011 at 4:49 PM, Ethan Rowe et...@the-rowes.com wrote: How do you plan to read the data? Entire histories, or in relatively confined slices of time? Do the events have any attributes by which you might segregate them, apart from time? If you can divide time into a fixed series of intervals, you can insert members of a given interval as columns (or supercolumns) in a row. But it depends how you want to use the data on the read side. On Thu, Apr 14, 2011 at 12:25 PM, Guillermo Winkler gwink...@inconcertcc.com wrote: I have a huge number of events I need to consume later, ordered by the date the event occured. My first approach to this problem was to use seconds since epoch as row key, and event ids as column names (empty value), this way: EventsByDate : { SecondsSinceEpoch: { evid:, evid:, evid: } } And use OPP as partitioner. Using GetRangeSlices to retrieve ordered events secuentially. Now I have two problems to solve: 1) The system is realtime, so all the events in a given moment are hitting the same box 2) Migrating from cassandra 0.6 to cassandra 0.7 OPP doesn't seem to like LongType for row keys, was this purposedly deprecated? I was thinking about secondary indexes, but it does not assure the order the rows are coming out of cassandra. Anyone has a better approach to model events by date given that restrictions? Thanks, Guille
Re: What's the best modeling approach for ordering events by date?
How do you plan to read the data? Entire histories, or in relatively confined slices of time? Do the events have any attributes by which you might segregate them, apart from time? If you can divide time into a fixed series of intervals, you can insert members of a given interval as columns (or supercolumns) in a row. But it depends how you want to use the data on the read side. On Thu, Apr 14, 2011 at 12:25 PM, Guillermo Winkler gwink...@inconcertcc.com wrote: I have a huge number of events I need to consume later, ordered by the date the event occured. My first approach to this problem was to use seconds since epoch as row key, and event ids as column names (empty value), this way: EventsByDate : { SecondsSinceEpoch: { evid:, evid:, evid: } } And use OPP as partitioner. Using GetRangeSlices to retrieve ordered events secuentially. Now I have two problems to solve: 1) The system is realtime, so all the events in a given moment are hitting the same box 2) Migrating from cassandra 0.6 to cassandra 0.7 OPP doesn't seem to like LongType for row keys, was this purposedly deprecated? I was thinking about secondary indexes, but it does not assure the order the rows are coming out of cassandra. Anyone has a better approach to model events by date given that restrictions? Thanks, Guille
hadoop streaming input
Hello. What's the current thinking on input support for Hadoop streaming? It seems like the relevant Jira issue has been quiet for some time: https://issues.apache.org/jira/browse/CASSANDRA-1497 Thanks. - Ethan
Re: hadoop streaming input
Thanks, Jeremy. I looked over the work that was done and it seemed like it was mostly there, though some comments in the ticket indicated possible problems. I may well need to take a crack at this sometime in the next few weeks, but if somebody beats me to it, I certainly won't complain. On Thu, Mar 17, 2011 at 2:06 PM, Jeremy Hanna jeremy.hanna1...@gmail.comwrote: I started it and added the tentative patch at the end of October. It needs to be rebased with the current 0.7-branch and completed - it's mostly there. I just tried to abstract some things in the process. I have changed jobs since then and I just haven't had time with the things I've been doing here. If you'd like to take a stab at it, you're welcome to rebase and get it finished. On Mar 17, 2011, at 12:57 PM, Ethan Rowe wrote: Hello. What's the current thinking on input support for Hadoop streaming? It seems like the relevant Jira issue has been quiet for some time: https://issues.apache.org/jira/browse/CASSANDRA-1497 Thanks. - Ethan
Re: ORM in Cassandra?
On 04/26/2010 01:26 PM, Isaac Arias wrote: On Apr 26, 2010, at 12:13 PM, Geoffry Roberts wrote: Clearly Cassandra is not an RDBMS. The intent of my Hibernate reference was to be more lyrical. Sorry if that didn't come through. Nonetheless, the need remains to relieve ourselves from excessive boilerplate coding. I agree with eliminating boilerplate code. Chris Shorrock wrote a simple object mapper in Scala for his Cascal Cassandra client. You may want to check out the wiki on GitHub (http://wiki.github.com/shorrockin/cascal/). In my opinion, a mapping solution for Cassandra should be more like a Template. Something that helps map (back and forth) rows to objects, columns to properties, etc. Since the data model can vary so much depending on data access patters, any overly structured approach that prescribes a particular schema will be of limited use. For what it's worth, this is exactly my opinion after looking at the problem for a bit, and I'm actively developing such a solution in Ruby. I spent some time playing with the CassandraObject project, but felt that despite all the good work that went in there, it didn't feel to me like it fit the problem space in an idiomatic manner. No criticism intended there; it seems to lean a little more towards a very structured schema, with less flexibility for things like collection attributes the members of which all have a key that matches a pattern (which is a use case we have). So, for my approach, there's one project that gives metaprogramming semantics for building the mapping behavior you describe: build classes that are oriented towards mapping between simple JSON-like structures and full-blown business objects. And a separate project that layers Cassandra specifics on top of that underlying mapper tool. The rub being: it's for a client, and we're collectively sorting out the details for releasing the code in some useful, public manner. But hopefully I'll get something useful out there for potential Ruby enthusiasts before too long. Hopefully a week or two. Thanks. - Ethan -- Ethan Rowe End Point Corporation et...@endpoint.com
Re: ORM in Cassandra?
On 04/26/2010 03:11 PM, Tatu Saloranta wrote: On Mon, Apr 26, 2010 at 10:35 AM, Ethan Roweet...@endpoint.com wrote: On 04/26/2010 01:26 PM, Isaac Arias wrote: On Apr 26, 2010, at 12:13 PM, Geoffry Roberts wrote: ... In my opinion, a mapping solution for Cassandra should be more like a Template. Something that helps map (back and forth) rows to objects, columns to properties, etc. Since the data model can vary so much depending on data access patters, any overly structured approach that prescribes a particular schema will be of limited use. For what it's worth, this is exactly my opinion after looking at the problem for a bit, and I'm actively developing such a solution in Ruby. I spent ... So, for my approach, there's one project that gives metaprogramming semantics for building the mapping behavior you describe: build classes that are oriented towards mapping between simple JSON-like structures and full-blown business objects. And a separate project that layers Cassandra specifics on top of that underlying mapper tool. +1 I think proper layering is the way to go: it makes problem (of simple construction of services that use Cassandra as the storage system) much easier to solve, divide and conquer. There are pretty decent OJM/OXM solutions that are mostly orthogonal wrt distributed storage part. I understand that there are some trade-offs (some things are easiest to optimize when Cassandra core handles them), but flexibility and best-tool-for-the-job have their benefits too. Right. Additionally, this mapping layer between simple (i.e. JSON-ready) structures and complex (i.e. business objects) would seem to be of much more general value than a Cassandra-specific mapper. I would think most any environment with a heavy reliance on Thrift services would benefit from such tools. -- Ethan Rowe End Point Corporation et...@endpoint.com
Re: Separate disks with cloud deployment
On 03/25/2010 11:10 AM, Mark Greene wrote: The FAQ page makes mention of using separate disks for the commit log and data directory. How would one go about achieving this in a cloud deployment such as Rackspace cloud servers or EC2 EBS? Or is it just preferred to use dedicated hardware to get the optimal performance? With EC2, you can mount more than one EBS device on a given server, so it's not a big deal. Mount one volume for logs, another volume for data. Additionally, we've found some benefit from running (for Postgres) RAID0 arrays on EBS; you get better I/O throughput. I'll defer to the Rackspace folks regarding Rackspace Cloud; it has been I/O on average since you're dealing with a real, local disk. But I don't know about getting a second disk in that environment, though. -- Ethan Rowe End Point Corporation et...@endpoint.com
Re: Separate disks with cloud deployment
On 03/25/2010 11:18 AM, Ethan Rowe wrote: [snip] I'll defer to the Rackspace folks regarding Rackspace Cloud; it has been I/O on average since you're dealing with a real, local disk. But I don't know about getting a second disk in that environment, though. That should have said better I/O on average. -- Ethan Rowe End Point Corporation et...@endpoint.com