[
https://issues.jboss.org/browse/ISPN-1658?page=com.atlassian.jira.plugin....
]
Dan Berindei reopened ISPN-1658:
--------------------------------
My fix caused another problem with cache view installation.
If a node has already left before the coordinator sends the PREPARE_VIEW command, the RPC
will fail without sending anything to any nodes.
If the coordinator then sends an APPLY_STATE command to another node, that command will
block waiting for the state transfer to start, but the coordinator won't roll back the
cache view installation and start a new one until the APPLY_STATE RPC times out.
RPCs invoked with SYNCRONOUS_IGNORE_LEAVERS taking too long if one of
the targets had already left
--------------------------------------------------------------------------------------------------
Key: ISPN-1658
URL:
https://issues.jboss.org/browse/ISPN-1658
Project: Infinispan
Issue Type: Bug
Components: RPC
Affects Versions: 5.1.0.CR2
Reporter: Dan Berindei
Assignee: Dan Berindei
Fix For: 5.1.0.CR3
Attachments: mlnt.txt
We were waiting for responses from the leavers as well, and then we were ignoring the
timeout. Instead we should remove the leavers from the targets list before sending the
command.
`SYNCHRONOUS_IGNORE_LEAVERS` is used for `CommitCommand`s and
`CacheViewControlCommand{COMMIT_VIEW}`s, if this happens it will block the cache view
installation (and so the entire cluster) until the RPC timeout expires.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.jboss.org/secure/ContactAdministrators!default.jspa
For more information on JIRA, see:
http://www.atlassian.com/software/jira