geode-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF subversion and git services (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (GEODE-7038) After auto-reconnect a server's multicat communications aren't working correctly
Date Wed, 31 Jul 2019 21:11:00 GMT

    [ https://issues.apache.org/jira/browse/GEODE-7038?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16897529#comment-16897529
] 

ASF subversion and git services commented on GEODE-7038:
--------------------------------------------------------

Commit 23727962252d5cbdf5dd7cccbd91ab65eca3ab19 in geode's branch refs/heads/feature/GEODE-7038
from Bruce Schuchardt
[ https://gitbox.apache.org/repos/asf?p=geode.git;h=2372796 ]

GEODE-7038: After auto-reconnect a server's multicat communications aren't working correctly

Ensure that a JoinResponseMessage is sent if multicast is enabled.  This
allows JGroupsMessenger to piggy-back a multicast message digest on the
response that the new process can install in its JGroups stack to ensure
that multicast messaging is properly initialized.

I've also replaced complex checks for whether UDP security is enabled
with a simpler check on ServiceConfig.  When UDP security is enabled we
are already sending a JoinResponseMessage and so we don't need to send
another one if multicast is enabled.


> After auto-reconnect a server's multicat communications aren't working correctly
> --------------------------------------------------------------------------------
>
>                 Key: GEODE-7038
>                 URL: https://issues.apache.org/jira/browse/GEODE-7038
>             Project: Geode
>          Issue Type: Bug
>          Components: membership, messaging
>            Reporter: Bruce Schuchardt
>            Assignee: Bruce Schuchardt
>            Priority: Major
>
> This was observed in an server having multicast enabled on a Region.  The server went
into a GC pause and was kicked out of the cluster.  After auto-reconnecting all of the servers
were requested to shut down and they all hung on destroy-region message responses.  Statistics
showed constant multicast retransmission requests but no retransmissions being sent.
> When a Region is configured to use multicast all of its cache operation messages are
multicast, including a destroy-region message.
> Some time ago we decided to stop sending Join Request Responses during discovery.  These
messages were responsible for carrying the JGroups multicast message digest so that a joining
member could install this digest into its multicast protocol.  Today these messages are only
sent if a UDP Diffie-Hellman algorithm has been specified.  We need to also ensure that we
send these messages if multicast is enabled.
>  



--
This message was sent by Atlassian JIRA
(v7.6.14#76016)

Mime
View raw message