cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Yuki Morishita (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-9882) DTCS (maybe other strategies) can block flushing when there are lots of sstables
Date Tue, 11 Aug 2015 21:14:33 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-9882?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14687393#comment-14687393
] 

Yuki Morishita commented on CASSANDRA-9882:
-------------------------------------------

As Sean's log above suggests, I think this can happen on other compaction strategy as well.
Compaction strategies try as much as possible in {{AbstractCompactionStrategy#getNextbackgroundTask}}
to create new compaction task with {{while(true)}} loop.
The method is {{synchronized}} so until compaction strategy find compacting sstables and mark
them compacting, other {{synchronized}} methods like {{handleNotification}} is blocked.

I suggest just removing {{while(true)}} loop from all compaction strategies and let them try
to find the next available candidates in the next background compaction round.
CASSANDRA-9592 keeps submitting background compaction so it should keep compaction continue
to work.

Patch here: https://github.com/yukim/cassandra/tree/9882-2

> DTCS (maybe other strategies) can block flushing when there are lots of sstables
> --------------------------------------------------------------------------------
>
>                 Key: CASSANDRA-9882
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-9882
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Core
>            Reporter: Jeremiah Jordan
>            Assignee: Marcus Eriksson
>              Labels: dtcs
>             Fix For: 2.1.x, 2.2.x
>
>
> MemtableFlushWriter tasks can get blocked by Compaction getNextBackgroundTask.  This
is in a wonky cluster with 200k sstables in the CF, but seems bad for flushing to be blocked
by getNextBackgroundTask when we are trying to make these new "smart" strategies that may
take some time to calculate what to do.
> {noformat}
> "MemtableFlushWriter:21" daemon prio=10 tid=0x00007ff7ad965000 nid=0x6693 waiting for
monitor entry [0x00007ff78a667000]
>    java.lang.Thread.State: BLOCKED (on object monitor)
> 	at org.apache.cassandra.db.compaction.WrappingCompactionStrategy.handleNotification(WrappingCompactionStrategy.java:237)
> 	- waiting to lock <0x00000006fcdbbf60> (a org.apache.cassandra.db.compaction.WrappingCompactionStrategy)
> 	at org.apache.cassandra.db.DataTracker.notifyAdded(DataTracker.java:518)
> 	at org.apache.cassandra.db.DataTracker.replaceFlushed(DataTracker.java:178)
> 	at org.apache.cassandra.db.compaction.AbstractCompactionStrategy.replaceFlushed(AbstractCompactionStrategy.java:234)
> 	at org.apache.cassandra.db.ColumnFamilyStore.replaceFlushed(ColumnFamilyStore.java:1475)
> 	at org.apache.cassandra.db.Memtable$FlushRunnable.runMayThrow(Memtable.java:336)
> 	at org.apache.cassandra.utils.WrappedRunnable.run(WrappedRunnable.java:28)
> 	at com.google.common.util.concurrent.MoreExecutors$SameThreadExecutorService.execute(MoreExecutors.java:297)
> 	at org.apache.cassandra.db.ColumnFamilyStore$Flush.run(ColumnFamilyStore.java:1127)
> 	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
> 	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
> 	at java.lang.Thread.run(Thread.java:745)
>    Locked ownable synchronizers:
> 	- <0x0000000743b3ac38> (a java.util.concurrent.ThreadPoolExecutor$Worker)
> "MemtableFlushWriter:19" daemon prio=10 tid=0x00007ff7ac57a000 nid=0x649b waiting for
monitor entry [0x00007ff78b8ee000]
>    java.lang.Thread.State: BLOCKED (on object monitor)
> 	at org.apache.cassandra.db.compaction.WrappingCompactionStrategy.handleNotification(WrappingCompactionStrategy.java:237)
> 	- waiting to lock <0x00000006fcdbbf60> (a org.apache.cassandra.db.compaction.WrappingCompactionStrategy)
> 	at org.apache.cassandra.db.DataTracker.notifyAdded(DataTracker.java:518)
> 	at org.apache.cassandra.db.DataTracker.replaceFlushed(DataTracker.java:178)
> 	at org.apache.cassandra.db.compaction.AbstractCompactionStrategy.replaceFlushed(AbstractCompactionStrategy.java:234)
> 	at org.apache.cassandra.db.ColumnFamilyStore.replaceFlushed(ColumnFamilyStore.java:1475)
> 	at org.apache.cassandra.db.Memtable$FlushRunnable.runMayThrow(Memtable.java:336)
> 	at org.apache.cassandra.utils.WrappedRunnable.run(WrappedRunnable.java:28)
> 	at com.google.common.util.concurrent.MoreExecutors$SameThreadExecutorService.execute(MoreExecutors.java:297)
> 	at org.apache.cassandra.db.ColumnFamilyStore$Flush.run(ColumnFamilyStore.java:1127)
> 	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
> 	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
> 	at java.lang.Thread.run(Thread.java:745)
> "CompactionExecutor:14" daemon prio=10 tid=0x00007ff7ad359800 nid=0x4d59 runnable [0x00007fecce3ea000]
>    java.lang.Thread.State: RUNNABLE
> 	at org.apache.cassandra.io.sstable.SSTableReader.equals(SSTableReader.java:628)
> 	at com.google.common.collect.ImmutableSet.construct(ImmutableSet.java:206)
> 	at com.google.common.collect.ImmutableSet.construct(ImmutableSet.java:220)
> 	at com.google.common.collect.ImmutableSet.access$000(ImmutableSet.java:74)
> 	at com.google.common.collect.ImmutableSet$Builder.build(ImmutableSet.java:531)
> 	at com.google.common.collect.Sets$1.immutableCopy(Sets.java:606)
> 	at org.apache.cassandra.db.ColumnFamilyStore.getOverlappingSSTables(ColumnFamilyStore.java:1352)
> 	at org.apache.cassandra.db.compaction.DateTieredCompactionStrategy.getNextBackgroundSSTables(DateTieredCompactionStrategy.java:88)
> 	at org.apache.cassandra.db.compaction.DateTieredCompactionStrategy.getNextBackgroundTask(DateTieredCompactionStrategy.java:65)
> 	- locked <0x00000006fcdbbf00> (a org.apache.cassandra.db.compaction.DateTieredCompactionStrategy)
> 	at org.apache.cassandra.db.compaction.WrappingCompactionStrategy.getNextBackgroundTask(WrappingCompactionStrategy.java:72)
> 	- locked <0x00000006fcdbbf60> (a org.apache.cassandra.db.compaction.WrappingCompactionStrategy)
> 	at org.apache.cassandra.db.compaction.CompactionManager$BackgroundCompactionTask.run(CompactionManager.java:238)
> 	at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:471)
> 	at java.util.concurrent.FutureTask.run(FutureTask.java:262)
> 	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
> 	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
> 	at java.lang.Thread.run(Thread.java:745)
> {noformat}



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message