[
https://issues.apache.org/jira/browse/CASSANDRA-12728?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15937459#comment-15937459
]
Garvit Juniwal commented on CASSANDRA-12728:
--------------------------------------------
[~jjirsa]
Assuming that writes are only acknowledged to client after hints are "synced"
to disk, I believe that patch I have is correct because you can safely ignore
any partially flushed hint at the end of file that was not synced.
If hints are written lazily (i.e., writes are acknowledged even before syncing
hints to disk,) it is implicit that Cassandra is resilient to loosing hints in
face of crashes. Even in this scenario, dropping the last partially written
hint is correct.
I do not understand the suggestion of making this an operator decision. There
could be hints that did not make it to disk in the of face of crashes, and I do
not know how you would detect them and force a crash.
> Handling partially written hint files
> -------------------------------------
>
> Key: CASSANDRA-12728
> URL: https://issues.apache.org/jira/browse/CASSANDRA-12728
> Project: Cassandra
> Issue Type: Bug
> Reporter: Sharvanath Pathak
> Labels: lhf
> Attachments: CASSANDRA-12728.patch
>
>
> {noformat}
> ERROR [HintsDispatcher:1] 2016-09-28 17:44:43,397
> HintsDispatchExecutor.java:225 - Failed to dispatch hints file
> d5d7257c-9f81-49b2-8633-6f9bda6e3dea-1474892654160-1.hints: file is corrupted
> ({})
> org.apache.cassandra.io.FSReadError: java.io.EOFException
> at
> org.apache.cassandra.hints.HintsReader$BuffersIterator.computeNext(HintsReader.java:282)
> ~[apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.hints.HintsReader$BuffersIterator.computeNext(HintsReader.java:252)
> ~[apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.utils.AbstractIterator.hasNext(AbstractIterator.java:47)
> ~[apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.hints.HintsDispatcher.sendHints(HintsDispatcher.java:156)
> ~[apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.hints.HintsDispatcher.sendHintsAndAwait(HintsDispatcher.java:137)
> ~[apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.hints.HintsDispatcher.dispatch(HintsDispatcher.java:119)
> ~[apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.hints.HintsDispatcher.dispatch(HintsDispatcher.java:91)
> ~[apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.hints.HintsDispatchExecutor$DispatchHintsTask.deliver(HintsDispatchExecutor.java:259)
> [apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.hints.HintsDispatchExecutor$DispatchHintsTask.dispatch(HintsDispatchExecutor.java:242)
> [apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.hints.HintsDispatchExecutor$DispatchHintsTask.dispatch(HintsDispatchExecutor.java:220)
> [apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.hints.HintsDispatchExecutor$DispatchHintsTask.run(HintsDispatchExecutor.java:199)
> [apache-cassandra-3.0.6.jar:3.0.6]
> at
> java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
> [na:1.8.0_77]
> at java.util.concurrent.FutureTask.run(FutureTask.java:266)
> [na:1.8.0_77]
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
> [na:1.8.0_77]
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
> [na:1.8.0_77]
> at java.lang.Thread.run(Thread.java:745) [na:1.8.0_77]
> Caused by: java.io.EOFException: null
> at
> org.apache.cassandra.io.util.RebufferingInputStream.readFully(RebufferingInputStream.java:68)
> ~[apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.io.util.RebufferingInputStream.readFully(RebufferingInputStream.java:60)
> ~[apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.hints.ChecksummedDataInput.readFully(ChecksummedDataInput.java:126)
> ~[apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.utils.ByteBufferUtil.read(ByteBufferUtil.java:402)
> ~[apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.hints.HintsReader$BuffersIterator.readBuffer(HintsReader.java:310)
> ~[apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.hints.HintsReader$BuffersIterator.computeNextInternal(HintsReader.java:301)
> ~[apache-cassandra-3.0.6.jar:3.0.6]
> at
> org.apache.cassandra.hints.HintsReader$BuffersIterator.computeNext(HintsReader.java:278)
> ~[apache-cassandra-3.0.6.jar:3.0.6]
> ... 15 common frames omitted
> {noformat}
> We've found out that the hint file was truncated because there was a hard
> reboot around the time of last write to the file. I think we basically need
> to handle partially written hint files. Also, the CRC file does not exist in
> this case (probably because it crashed while writing the hints file). May be
> ignoring and cleaning up such partially written hint files can be a way to
> fix this?
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
