[
https://issues.apache.org/jira/browse/IGNITE-10926?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16742260#comment-16742260
]
Amelchev Nikita commented on IGNITE-10926:
------------------------------------------
I have prepared [PR|https://github.com/apache/ignite/pull/5820] with reproducer
to fix the issue. I used the watcher for the local alive node. It resolve case
when the node was deleted after several cluster restarts.
Another way for resolving the issue was using a special event to invoke
reconnect. But this requires creating a new discovery event because a custom
event will not be processed (rtState isn't joined).
[~sergey-chugunov], could you take a look at the issue, please?
> ZookeeperDiscoverySpi: client does not survive after several cluster restarts
> -----------------------------------------------------------------------------
>
> Key: IGNITE-10926
> URL: https://issues.apache.org/jira/browse/IGNITE-10926
> Project: Ignite
> Issue Type: Bug
> Components: zookeeper
> Reporter: Amelchev Nikita
> Assignee: Amelchev Nikita
> Priority: Major
> Fix For: 2.8
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> {{ZookeeperDiscoveryImpl#cleanupPreviousClusterData}} can delete alive node
> of a client in case of low internal order.
> Steps to reproduce:
> 1. Start server and client.
> 2. Stop the server and wait for the client disconnected.
> 3. Start and stop the server. The server hasn't time to process client join
> request.
> 4. Start server. It will delete alive client node because the client has low
> internal order. The client will never connect.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
