[
https://issues.jboss.org/browse/ISPN-7051?page=com.atlassian.jira.plugin....
]
Radim Vansa commented on ISPN-7051:
-----------------------------------
So, what happens:
1. C does a staggered get to B, A (current topologyId=4)
2. B responds first with SuccessfulResponse{null} - I am not sure why it does not have the
value if it's an owner according to readCH
3. A responds second with actual value, but the delayed response is ignored
4. C retries to find the entry, routing accoring to pendingCH - and the owner there is C
5. C sends the command to itself, but as local node is filtered out in JGroupsTransport,
it returns empty response map
A correct solution should involve a check if the actual owners - self isn't empty (or
even better, forcing to send remote commands with remote recipients only), but the logic
is changed in
https://github.com/infinispan/infinispan/pull/4564 so I am not sure if we
should invest more into this, rather than do a review & merge on the above mentioned
PR.
Intermittent RpcException during PUTs and GETs when topology changes
--------------------------------------------------------------------
Key: ISPN-7051
URL:
https://issues.jboss.org/browse/ISPN-7051
Project: Infinispan
Issue Type: Bug
Components: Core
Reporter: Gustavo Fernandes
Priority: Critical
Attachments: trace-jgroups.tgz.part-00, trace-jgroups.tgz.part-01,
trace-jgroups.tgz.part-02, trace-jgroups.tgz.part-03, trace-jgroups.tgz.part-04,
trace.tgz
The commit
https://github.com/infinispan/infinispan/commit/6ae4e2e7707285287ef753661...
seems to be introducing intermittent errors during gets and puts under topology changes
--
This message was sent by Atlassian JIRA
(v6.4.11#64026)