[
https://issues.apache.org/jira/browse/CASSANDRA-15263?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16904954#comment-16904954
]
Benedict commented on CASSANDRA-15263:
--------------------------------------
Thanks [[email protected]]. I can see what's happening now, and it
looks benign. It should resolve when you finish upgrading the nodes in your
cluster.
The error is caused by the rare scenario of the rows not using all of the
declared clustering columns, which inserts a {{null}} clustering value for
{{column2}}. It was not thought by the author of the legacy converter that a
RT clustering component could be {{null}}, and they would have ordinarily been
correct as row deletions are no longer stored as range tombstones in 3.0,
however synthetic range tombstone bounds can be built from row clusterings, and
since the row has a null component, the synthetic RT does also.
The fix for 3.0 would be simple, namely to ignore the {{null}} value when
computing a digest, however it looks like this {{null}} is also incompatible
with 2.1, since it could legitimately never arise there, without the new
machinery of 3.0 that synthesises them. So sending this synthetic clustering
to a 2.1 node could be more harmful than throwing this exception.
I will have to think about the best recourse to address this in 3.0 without
adversely impacting a 2.1 node.
> LegacyLayout RangeTombstoneList throws java.lang.NullPointerException: null
> ---------------------------------------------------------------------------
>
> Key: CASSANDRA-15263
> URL: https://issues.apache.org/jira/browse/CASSANDRA-15263
> Project: Cassandra
> Issue Type: Bug
> Components: Cluster/Schema
> Reporter: feroz shaik
> Assignee: Benedict
> Priority: Normal
> Labels: 2.1.16, 3.11.4
> Attachments: sample.system.log, schema.txt,
> sstabledump_sal_purge_d03.json, sstablemetadata_sal_purge_d03,
> stack_trace.txt, system.log, system.log, system.log, system.log,
> system_latest.log
>
>
> We have hit a problem today while upgrading from 2.1.16 to 3.11.4.
> we encountered this as soon as the first node started up with 3.11.4
> The full error stack is attached - [^stack_trace.txt]
>
> The below errors continued in the log file as long as the process was up.
> ERROR [Native-Transport-Requests-12] 2019-08-06 03:00:47,135
> ErrorMessage.java:384 - Unexpected exception during request
> java.lang.NullPointerException: null
> ERROR [Native-Transport-Requests-8] 2019-08-06 03:00:48,778
> ErrorMessage.java:384 - Unexpected exception during request
> java.lang.NullPointerException: null
> ERROR [Native-Transport-Requests-13] 2019-08-06 03:00:57,454
>
> The nodetool version says 3.11.4 and the no of connections on native por t-
> 9042 was similar to other nodes. The exceptions were scary that we had to
> call off the change. Any help and insights to this problem from the community
> is appreciated.
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
