[
https://issues.apache.org/jira/browse/CASSANDRA-5804?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13719861#comment-13719861
]
Ashley Winters commented on CASSANDRA-5804:
-------------------------------------------
The dead connections are cross-datacenter, and I believe I have some network
equipment _(naughty firewall)_ resetting idle connections. That'd be why gossip
says the node is up, even though a pooled {{OutboundTcpConnection}} would be
reset due to idleness.
> AntiEntropySession fails when OutboundTcpConnection receives IOException
> ------------------------------------------------------------------------
>
> Key: CASSANDRA-5804
> URL: https://issues.apache.org/jira/browse/CASSANDRA-5804
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Affects Versions: 1.2.6
> Environment: CentOS 6.3 x86_64
> Reporter: Ashley Winters
> Attachments: ioexception.txt
>
>
> When requesting merkle trees for a repair, if the OutboundTcpConnection
> grabbed from the connection pool is reset (java.io.IOException: Connection
> reset by peer), the target node is not marked as dead, the TREE_REQUEST is
> not retried, and the repair does not fail. Instead, the repair stalls waiting
> for the merkle tree response which will never arrive.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
