accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Daniel Blum (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ACCUMULO-4435) WholeColumnFamilyIterator doesn't work after re-init
Date Wed, 31 Aug 2016 17:41:21 GMT

    [ https://issues.apache.org/jira/browse/ACCUMULO-4435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15452851#comment-15452851
] 

Daniel Blum commented on ACCUMULO-4435:
---------------------------------------

It should be easily reproducible - just use WCFI on any scan that will last long enough to
require re-initialization (I don't know how Accumulo decides when to do this).

I'm afraid we don't have any budget to spend on patch creation, since using WCFI is not critical
to us.

> WholeColumnFamilyIterator doesn't work after re-init
> ----------------------------------------------------
>
>                 Key: ACCUMULO-4435
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-4435
>             Project: Accumulo
>          Issue Type: Bug
>          Components: core
>    Affects Versions: 1.6.5
>            Reporter: Daniel Blum
>            Priority: Minor
>              Labels: iterators
>
> When using WholeColumnFamilyIterator this exception is thrown:
> {noformat}
> java.lang.IllegalStateException: never been seeked
>         at org.apache.accumulo.core.iterators.WrappingIterator.hasTop(WrappingIterator.java:80)
>         at org.apache.accumulo.core.iterators.user.WholeColumnFamilyIterator.hasTop(WholeColumnFamilyIterator.java:213)
>         at org.apache.accumulo.core.iterators.system.SourceSwitchingIterator.readNext(SourceSwitchingIterator.java:132)
>         at org.apache.accumulo.core.iterators.system.SourceSwitchingIterator.seek(SourceSwitchingIterator.java:167)
>         at org.apache.accumulo.tserver.Tablet.lookup(Tablet.java:1523)
>         at org.apache.accumulo.tserver.Tablet.lookup(Tablet.java:1630)
>         at org.apache.accumulo.tserver.TabletServer$ThriftClientHandler$LookupTask.run(TabletServer.java:1155)
>         at org.apache.accumulo.trace.instrument.TraceRunnable.run(TraceRunnable.java:47)
>         at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
>         at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
>         at org.apache.accumulo.trace.instrument.TraceRunnable.run(TraceRunnable.java:47)
>         at org.apache.accumulo.core.util.LoggingRunnable.run(LoggingRunnable.java:34)
>         at java.lang.Thread.run(Thread.java:745)
> {noformat}
> I THINK this is caused by re-initialization of the iterator stack. When this happens
WCFI.seek() may end up returning without calling seek() on its source (because it gets a range
with effectively nothing in it). This would be fine except that WCFI.hasTop() calls hasTop()
on its source; since the source has not been seeked yet, the exception is thrown.
> WholeRowIterator does not run into this problem in spite of using the same seek logic
because its hasTop (actually RowEncodingIterator.hasTop) does not check the source's hasTop.
(The way RowEncodingIterator and WCFI handle all of this does not really conform to the interface
definition - they should really keep track themselves of whether they have been seeked and
throw the appropriate exceptions.)
> I actually encountered this in 1.6.2, but I can't select that as a version here. I believe
the problem also exists in 1.6.5 (and 1.7.2) based on inspecting the source, but cannot test
with those versions right now to confirm.



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

Mime
View raw message