You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Andrew Hust (JIRA)" <ji...@apache.org> on 2015/09/30 22:22:04 UTC
[jira] [Comment Edited] (CASSANDRA-10360)
UnsupportedOperationException when compacting system.size_estimates after
2.1 -> 2.2 -> 3.0 upgrade
[ https://issues.apache.org/jira/browse/CASSANDRA-10360?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14937461#comment-14937461 ]
Andrew Hust edited comment on CASSANDRA-10360 at 9/30/15 8:21 PM:
------------------------------------------------------------------
I have now reproduced this issue when upgrading from 2.1 to 3.0 when doing a rolling upgrade with cstar_perf. Attached site_estimates sstables and system log from one of the nodes. I'll work on getting a isolated script to reproduce.
UPDATE: The above shell script has been updated to show failure from 2.1 -> 3.0. It just needed more activity on the stress operation to trigger the issue. Change line {{for cver in 3.0}} to {{for cver in 2.2 3.0}} for the original upgrade path.
was (Author: nutbunnies):
I have now reproduced this issue when upgrading from 2.1 to 3.0 when doing a rolling upgrade with cstar_perf. Attached site_estimates sstables and system log from one of the nodes. I'll work on getting a isolated script to reproduce.
> UnsupportedOperationException when compacting system.size_estimates after 2.1 -> 2.2 -> 3.0 upgrade
> ---------------------------------------------------------------------------------------------------
>
> Key: CASSANDRA-10360
> URL: https://issues.apache.org/jira/browse/CASSANDRA-10360
> Project: Cassandra
> Issue Type: Bug
> Reporter: Andrew Hust
> Assignee: Sylvain Lebresne
> Priority: Blocker
> Fix For: 3.0.0 rc2
>
> Attachments: size_estimates.tar.bz2, system.log.bz2
>
>
> When upgrading from 2.1 -> 2.2 -> 3.0 the system.size_estimates table will get stuck in a compaction loop throwing:
> {code}
> java.lang.UnsupportedOperationException
> at org.apache.cassandra.db.UnfilteredDeserializer$OldFormatDeserializer.readNext(UnfilteredDeserializer.java:382)
> at org.apache.cassandra.io.sstable.SSTableSimpleIterator$OldFormatIterator.computeNext(SSTableSimpleIterator.java:147)
> at org.apache.cassandra.io.sstable.SSTableSimpleIterator$OldFormatIterator.computeNext(SSTableSimpleIterator.java:96)
> at org.apache.cassandra.utils.AbstractIterator.hasNext(AbstractIterator.java:47)
> at org.apache.cassandra.io.sstable.SSTableIdentityIterator.computeNext(SSTableIdentityIterator.java:100)
> at org.apache.cassandra.io.sstable.SSTableIdentityIterator.computeNext(SSTableIdentityIterator.java:30)
> at org.apache.cassandra.utils.AbstractIterator.hasNext(AbstractIterator.java:47)
> at org.apache.cassandra.db.rows.LazilyInitializedUnfilteredRowIterator.computeNext(LazilyInitializedUnfilteredRowIterator.java:95)
> at org.apache.cassandra.db.rows.LazilyInitializedUnfilteredRowIterator.computeNext(LazilyInitializedUnfilteredRowIterator.java:32)
> at org.apache.cassandra.utils.AbstractIterator.hasNext(AbstractIterator.java:47)
> at org.apache.cassandra.utils.MergeIterator$TrivialOneToOne.computeNext(MergeIterator.java:460)
> at org.apache.cassandra.utils.AbstractIterator.hasNext(AbstractIterator.java:47)
> at org.apache.cassandra.db.rows.UnfilteredRowIterators$UnfilteredRowMergeIterator.computeNext(UnfilteredRowIterators.java:503)
> at org.apache.cassandra.db.rows.UnfilteredRowIterators$UnfilteredRowMergeIterator.computeNext(UnfilteredRowIterators.java:363)
> at org.apache.cassandra.utils.AbstractIterator.hasNext(AbstractIterator.java:47)
> at org.apache.cassandra.db.rows.WrappingUnfilteredRowIterator.hasNext(WrappingUnfilteredRowIterator.java:80)
> at org.apache.cassandra.db.rows.AlteringUnfilteredRowIterator.hasNext(AlteringUnfilteredRowIterator.java:72)
> at org.apache.cassandra.db.rows.UnfilteredRowIterator.isEmpty(UnfilteredRowIterator.java:100)
> at org.apache.cassandra.db.partitions.PurgingPartitionIterator.hasNext(PurgingPartitionIterator.java:72)
> at org.apache.cassandra.db.compaction.CompactionIterator.hasNext(CompactionIterator.java:223)
> at org.apache.cassandra.db.compaction.CompactionTask.runMayThrow(CompactionTask.java:177)
> at org.apache.cassandra.utils.WrappedRunnable.run(WrappedRunnable.java:28)
> at org.apache.cassandra.db.compaction.CompactionTask.executeInternal(CompactionTask.java:78)
> at org.apache.cassandra.db.compaction.AbstractCompactionTask.execute(AbstractCompactionTask.java:60)
> at org.apache.cassandra.db.compaction.CompactionManager$8.runMayThrow(CompactionManager.java:539)
> at org.apache.cassandra.utils.WrappedRunnable.run(WrappedRunnable.java:28)
> at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
> at java.util.concurrent.FutureTask.run(FutureTask.java:266)
> at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
> at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
> at java.lang.Thread.run(Thread.java:745)
> {code}
> It will only occur when upgrading thru 2.2. Going from 2.1 -> 3.0 will not surface the issue. It can be reproduced with the following (note -- changing jdks will need to be altered if you aren't on OSX)
> {code}
> #!/bin/sh
> echo "using java7 for cassandra-2.1 instance"
> export JAVA_HOME=$(/usr/libexec/java_home -v 1.7)
> ccm stop
> ccm remove upgrade_nodes
> ccm create -n 1 -v git:cassandra-2.1 upgrade_nodes
> ccm start
> ccm node1 stress write n=500K -rate threads=4 -mode native cql3
> sleep 10
> for cver in 2.2 3.0
> do
> echo "Draining all nodes"
> ccm node1 nodetool drain
> ccm stop
> echo "switching to java 8"
> export JAVA_HOME=$(/usr/libexec/java_home -v 1.8)
> echo "Upgrading to git:cassandra-$cver"
> ccm setdir -v git:cassandra-$cver
> ccm start
> echo "Sleeping to all version migrations"
> sleep 30
> echo "Upgrading sstables"
> ccm node1 nodetool upgradesstables
> ccm node1 nodetool upgradesstables system
> ccm node1 nodetool compact system
> ccm node1 stress write n=500K -rate threads=4 -mode native cql3
> sleep 10
> done
> echo "***********"
> echo "instead of creating churn to cause compaction naturally forcing compaction of system keyspace"
> echo "***********"
> ccm node1 nodetool compact system
> ccm stop
> {code}
> The example uses {{nodetool compact system}} but it will also occur with {{nodetool upgradesstables system}}. I'm puzzled by that since the script runs {{upgradesstables}} on each iteration. Is the system keyspace not effected by the command without arguments?
> Ran against:
> 2.1: {{e889ee408bec5330c312ff6b72a81a0012fdf2a5}}
> 2.2: {{e63dacf793fedc8a9eed9c7fc635cde5f9fd68f3}}
> 3.0: {{9218d7456b36d20ebe78bab23594e67d2f0c4a20}}
> //CC [~enigmacurry]
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)