[
https://issues.apache.org/jira/browse/IGNITE-647?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14482176#comment-14482176
]
Alexey Goncharuk edited comment on IGNITE-647 at 4/6/15 11:20 PM:
------------------------------------------------------------------
Initial investigation results:
# Node 1 is started without cache "A" configuration. Start is not finished yet,
onKernalStart() is not called.
# Node 2 is started with cache "A" configuration, configured with fair affinity
function.
# During discovery data exchange node 1 collects info about cache "A" and adds
it to registered caches.
# Node 1 continues start process, iterates over registered caches and adds
cache "A" to the list of cache contexts.
# Node 1 initiates partition map exchange. Since cache "A" is already in the
list of started caches, node 1 calculates nodes for cache "A" on topology
version 1, but on topology version 1 cache "A" did not exist. This causes the
affinity fetch future hang.
Proposed solution: add topology version to discovery exchange process so that
local node does not start caches received from nodes started after it (steps 3,
4).
was (Author: agoncharuk):
Initial investigation results:
# Node 1 is started without cache "A" configuration. Start is not finished yet,
onKernalStart() is not called.
# Node 2 is started with cache "A" configuration, configured with fair affinity
function.
# During discovery data exchange node 1 collects info about cache "A" and adds
it to registered caches.
# Node 1 continues start process, iterates over registered caches and adds
cache "A" to the list of cache contexts.
# Node 1 initiates partition map exchange. Since cache "A" is already in the
list of started caches, node 1 calculates nodes for cache "A" on topology
version 1, but on topology version 1 cache "A" did not exist. This causes the
affinity fetch future hang.
Proposed solution: add topology version to discovery exchange process so that
local node does not start caches received from nodes started after it.
> org.apache.ignite.IgniteCacheAffinitySelfTest.testAffinity() hangs
> ------------------------------------------------------------------
>
> Key: IGNITE-647
> URL: https://issues.apache.org/jira/browse/IGNITE-647
> Project: Ignite
> Issue Type: Bug
> Reporter: Yakov Zhdanov
> Assignee: Alexey Goncharuk
> Priority: Blocker
> Fix For: sprint-4
>
> Attachments: threaddump.txt
>
>
> 1-2 runs out of ~10 local runs hanged for me
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
