[
https://issues.apache.org/jira/browse/IGNITE-5272?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16044267#comment-16044267
]
Semen Boikov commented on IGNITE-5272:
--------------------------------------
If all nodes are not notified synchronously about client cache start, then
there are are some issues with near cache support:
- need do changes in ATOMIC cache update protocol: it is possible that client
knows affinity mapping, but on primary there is info about near cache readers ,
need make sure client does not finish operation before readers are updated
- in tx cache need handle case when tx was marked as onePhase, but on primary
there is info about readers, in this case need switch back to non-onePhase mode
- if reader for client near cache was added and after that client cache was
closed, then reader update message won't be handled
> Try get rid of custom discovery message for client cache start
> --------------------------------------------------------------
>
> Key: IGNITE-5272
> URL: https://issues.apache.org/jira/browse/IGNITE-5272
> Project: Ignite
> Issue Type: Task
> Components: cache
> Reporter: Semen Boikov
> Assignee: Semen Boikov
> Priority: Critical
> Labels: important
> Fix For: 2.1
>
>
> Currently both server and client caches are started in the same way - using
> custom discovery message. Most probably for client caches we do not need
> strong ordering guarentees which discovery messages provide.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
