[
https://issues.apache.org/jira/browse/CASSANDRA-1903?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12975359#action_12975359
]
Eric Evans commented on CASSANDRA-1903:
---------------------------------------
Upon further examination...
The null value here is row.cf, it's null because the read came up empty, so
this looks like a race to me (the read is beating the write). But... the
bigger question is, why have we added a read to the write path? This seems very
wrong to me.
Also, what does ReplicateOnWriteTask do that the existing replication doesn't?
How come we're doing both? This all seems to have changed as part of the
counters implementation, so it would be great if someone involved with that
could sound off on this.
> NullPointerException from o.a.c.db.ReplicateOnWriteTask
> -------------------------------------------------------
>
> Key: CASSANDRA-1903
> URL: https://issues.apache.org/jira/browse/CASSANDRA-1903
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Reporter: Eric Evans
> Attachments: thrift-test.py
>
>
> I'm seeing a whole lot of these when writing to a node.
>
> {noformat}
> java.lang.NullPointerException
> at
> org.apache.cassandra.db.ReplicateOnWriteTask.run(ReplicateOnWriteTask.java:97)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1110)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:603)
> at java.lang.Thread.run(Thread.java:636)
> {noformat}
> I don't think it will be difficult to reproduce, but the script I'm using is
> attached.
> I bisected the source tree and and
> http://svn.apache.org/viewvc?view=rev&revision=1052356 seems to the culprit
> (a merge from 0.7). Maybe CASSANDRA-1530?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
