New node has high network and disk usage.
Hi, We recently added a new node to our cluster in order to replace a node that died (hardware failure we believe). For the next two weeks it had high disk and network activity. We replaced the server, but it's happened again. We've looked into memory allowances, disk performance, number of connections, and all the nodetool stats, but can't find the cause of the issue. `nodetool tpstats`[0] shows a lot of active and pending threads, in comparison to the rest of the cluster, but that's likely a symptom, not a cause. `nodetool status`[1] shows the cluster isn't quite balanced. The bad node (D) has less data. Disk Activity[2] and Network activity[3] on this node is far higher than the rest. The only other difference this node has to the rest of the cluster is that its on the ext4 filesystem, whereas the rest are ext3, but we've done plenty of testing there and can't see how that would affect performance on this node so much. Nothing of note in system.log. What should our next step be in trying to diagnose this issue? Best wishes, Vic [0] `nodetool tpstats` output: Good node: Pool NameActive Pending Completed Blocked All time blocked ReadStage 0 0 46311521 0 0 RequestResponseStage 0 0 23817366 0 0 MutationStage 0 0 47389269 0 0 ReadRepairStage 0 0 11108 0 0 ReplicateOnWriteStage 0 0 0 0 0 GossipStage 0 05259908 0 0 CacheCleanupExecutor 0 0 0 0 0 MigrationStage0 0 30 0 0 MemoryMeter 0 0 16563 0 0 FlushWriter 0 0 39637 026 ValidationExecutor0 0 19013 0 0 InternalResponseStage 0 0 9 0 0 AntiEntropyStage 0 0 38026 0 0 MemtablePostFlusher 0 0 81740 0 0 MiscStage 0 0 19196 0 0 PendingRangeCalculator0 0 23 0 0 CompactionExecutor0 0 61629 0 0 commitlog_archiver0 0 0 0 0 HintedHandoff 0 0 63 0 0 Message type Dropped RANGE_SLICE 0 READ_REPAIR 0 PAGED_RANGE 0 BINARY 0 READ 640 MUTATION 0 _TRACE 0 REQUEST_RESPONSE 0 COUNTER_MUTATION 0 Bad node: Pool NameActive Pending Completed Blocked All time blocked ReadStage32 113 52216 0 0 RequestResponseStage 0 0 4167 0 0 MutationStage 0 0 127559 0 0 ReadRepairStage 0 0125 0 0 ReplicateOnWriteStage 0 0 0 0 0 GossipStage 0 0 9965 0 0 CacheCleanupExecutor 0 0 0 0 0 MigrationStage0 0 0 0 0 MemoryMeter 0 0 24 0 0 FlushWriter 0 0 27 0 1 ValidationExecutor0 0 0 0 0 InternalResponseStage 0 0 0 0 0 AntiEntropyStage 0 0 0 0 0 MemtablePostFlusher 0 0 96 0 0 MiscStage 0 0 0 0 0 PendingRangeCalculator0 0 10 0 0 CompactionExecutor1 1 73 0 0 commitlog_archiver0 0 0 0 0 HintedHandoff 0 0 15 0 0 Message type Dropped RANGE_SLICE130 READ_REPAIR 1 PAGED_RANGE
Re: New node has high network and disk usage.
I should probably have mentioned that we're on Cassandra 2.0.10. On 6 January 2016 at 15:26, Vickrum Loi <vickrum@idioplatform.com> wrote: > Hi, > > We recently added a new node to our cluster in order to replace a node > that died (hardware failure we believe). For the next two weeks it had high > disk and network activity. We replaced the server, but it's happened again. > We've looked into memory allowances, disk performance, number of > connections, and all the nodetool stats, but can't find the cause of the > issue. > > `nodetool tpstats`[0] shows a lot of active and pending threads, in > comparison to the rest of the cluster, but that's likely a symptom, not a > cause. > > `nodetool status`[1] shows the cluster isn't quite balanced. The bad node > (D) has less data. > > Disk Activity[2] and Network activity[3] on this node is far higher than > the rest. > > The only other difference this node has to the rest of the cluster is that > its on the ext4 filesystem, whereas the rest are ext3, but we've done > plenty of testing there and can't see how that would affect performance on > this node so much. > > Nothing of note in system.log. > > What should our next step be in trying to diagnose this issue? > > Best wishes, > Vic > > [0] `nodetool tpstats` output: > > Good node: > Pool NameActive Pending Completed > Blocked All time blocked > ReadStage 0 0 46311521 > 0 0 > RequestResponseStage 0 0 23817366 > 0 0 > MutationStage 0 0 47389269 > 0 0 > ReadRepairStage 0 0 11108 > 0 0 > ReplicateOnWriteStage 0 0 0 > 0 0 > GossipStage 0 05259908 > 0 0 > CacheCleanupExecutor 0 0 0 > 0 0 > MigrationStage0 0 30 > 0 0 > MemoryMeter 0 0 16563 > 0 0 > FlushWriter 0 0 39637 > 026 > ValidationExecutor0 0 19013 > 0 0 > InternalResponseStage 0 0 9 > 0 0 > AntiEntropyStage 0 0 38026 > 0 0 > MemtablePostFlusher 0 0 81740 > 0 0 > MiscStage 0 0 19196 > 0 0 > PendingRangeCalculator0 0 23 > 0 0 > CompactionExecutor0 0 61629 > 0 0 > commitlog_archiver0 0 0 > 0 0 > HintedHandoff 0 0 63 > 0 0 > > Message type Dropped > RANGE_SLICE 0 > READ_REPAIR 0 > PAGED_RANGE 0 > BINARY 0 > READ 640 > MUTATION 0 > _TRACE 0 > REQUEST_RESPONSE 0 > COUNTER_MUTATION 0 > > Bad node: > Pool NameActive Pending Completed > Blocked All time blocked > ReadStage32 113 52216 > 0 0 > RequestResponseStage 0 0 4167 > 0 0 > MutationStage 0 0 127559 > 0 0 > ReadRepairStage 0 0125 > 0 0 > ReplicateOnWriteStage 0 0 0 > 0 0 > GossipStage 0 0 9965 > 0 0 > CacheCleanupExecutor 0 0 0 > 0 0 > MigrationStage0 0 0 > 0 0 > MemoryMeter 0 0 24 > 0 0 > FlushWriter 0 0 27 > 0 1 > ValidationExecutor0 0 0 > 0 0 > InternalResponseStage 0 0 0 > 0 0 > A
Re: New node has high network and disk usage.
# nodetool compactionstats pending tasks: 22 compaction typekeyspace table completed total unit progress Compactionproduction_analyticsinteractions 240410213161172668724 bytes 0.15% Compactionproduction_decisionsdecisions.decisions_q_idx 120815385 226295183 bytes53.39% Active compaction remaining time : 2h39m58s Worth mentioning that compactions haven't been running on this node particularly often. The node's been performing badly regardless of whether it's compacting or not. On 6 January 2016 at 16:35, Jeff Ferland <j...@tubularlabs.com> wrote: > What’s your output of `nodetool compactionstats`? > > On Jan 6, 2016, at 7:26 AM, Vickrum Loi <vickrum@idioplatform.com> > wrote: > > Hi, > > We recently added a new node to our cluster in order to replace a node > that died (hardware failure we believe). For the next two weeks it had high > disk and network activity. We replaced the server, but it's happened again. > We've looked into memory allowances, disk performance, number of > connections, and all the nodetool stats, but can't find the cause of the > issue. > > `nodetool tpstats`[0] shows a lot of active and pending threads, in > comparison to the rest of the cluster, but that's likely a symptom, not a > cause. > > `nodetool status`[1] shows the cluster isn't quite balanced. The bad node > (D) has less data. > > Disk Activity[2] and Network activity[3] on this node is far higher than > the rest. > > The only other difference this node has to the rest of the cluster is that > its on the ext4 filesystem, whereas the rest are ext3, but we've done > plenty of testing there and can't see how that would affect performance on > this node so much. > > Nothing of note in system.log. > > What should our next step be in trying to diagnose this issue? > > Best wishes, > Vic > > [0] `nodetool tpstats` output: > > Good node: > Pool NameActive Pending Completed > Blocked All time blocked > ReadStage 0 0 46311521 > 0 0 > RequestResponseStage 0 0 23817366 > 0 0 > MutationStage 0 0 47389269 > 0 0 > ReadRepairStage 0 0 11108 > 0 0 > ReplicateOnWriteStage 0 0 0 > 0 0 > GossipStage 0 05259908 > 0 0 > CacheCleanupExecutor 0 0 0 > 0 0 > MigrationStage0 0 30 > 0 0 > MemoryMeter 0 0 16563 > 0 0 > FlushWriter 0 0 39637 > 026 > ValidationExecutor0 0 19013 > 0 0 > InternalResponseStage 0 0 9 > 0 0 > AntiEntropyStage 0 0 38026 > 0 0 > MemtablePostFlusher 0 0 81740 > 0 0 > MiscStage 0 0 19196 > 0 0 > PendingRangeCalculator0 0 23 > 0 0 > CompactionExecutor0 0 61629 > 0 0 > commitlog_archiver0 0 0 > 0 0 > HintedHandoff 0 0 63 > 0 0 > > Message type Dropped > RANGE_SLICE 0 > READ_REPAIR 0 > PAGED_RANGE 0 > BINARY 0 > READ 640 > MUTATION 0 > _TRACE 0 > REQUEST_RESPONSE 0 > COUNTER_MUTATION 0 > > Bad node: > Pool NameActive Pending Completed > Blocked All time blocked > ReadStage32 113 52216 > 0 0 > RequestResponseStage 0 0 4167 > 0 0 > MutationStage 0 0 127559 > 0 0 > ReadRepairStage 0 0125 > 0 0 > ReplicateOnWriteStage 0 0 0 > 0 0 > Goss