[
https://issues.apache.org/jira/browse/CASSANDRA-8903?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14567094#comment-14567094
]
Sylvain Lebresne commented on CASSANDRA-8903:
---------------------------------------------
Getting the actual query that produce this would be really helpful as a priori
I can't find any case that would trigger this: The error implies that the
filter passed to {{sliceFilterToSC}} has 1) more than one slice and 2) is
reversed. Except that from Thrift, as far as I can tell, we never generate such
filter. We do generate multiple slices in one case, which is for super column
names filter, but we very explicitly set the filter 'not reversed' in that case.
> Super Columns exception during upgrade from C* 1.2.18 to C* 2.0.11
> ------------------------------------------------------------------
>
> Key: CASSANDRA-8903
> URL: https://issues.apache.org/jira/browse/CASSANDRA-8903
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Reporter: Erick Ramirez
> Assignee: Sylvain Lebresne
>
> While in the middle of an upgrade of a 12-node cluster, the following errors
> were constantly being reported in the logs:
> {noformat}
> ERROR [WRITE-/10.73.73.26] 2015-03-02 22:59:12,523 OutboundTcpConnection.java
> (line 234) error writing to /xx.xx.xx.xx
> java.lang.RuntimeException: Cannot convert filter to old super column format.
> Update all nodes to Cassandra 2.0 first.
> at
> org.apache.cassandra.db.SuperColumns.sliceFilterToSC(SuperColumns.java:353)
> at org.apache.cassandra.db.SuperColumns.filterToSC(SuperColumns.java:258)
> at
> org.apache.cassandra.db.RangeSliceCommandSerializer.serializedSize(RangeSliceCommand.java:284)
>
> at
> org.apache.cassandra.db.RangeSliceCommandSerializer.serializedSize(RangeSliceCommand.java:156)
>
> at org.apache.cassandra.net.MessageOut.serialize(MessageOut.java:116)
> at
> org.apache.cassandra.net.OutboundTcpConnection.writeInternal(OutboundTcpConnection.java:251)
>
> at
> org.apache.cassandra.net.OutboundTcpConnection.writeConnected(OutboundTcpConnection.java:203)
>
> at
> org.apache.cassandra.net.OutboundTcpConnection.run(OutboundTcpConnection.java:151)
> {noformat}
> Confirmed that code is accessing cluster via Thrift and not CQL API.
> This issue is very similar to CASSANDRA-6966 but may relate to the way that
> single slice and reversed is being handled.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
