[
https://issues.apache.org/jira/browse/GEODE-8656?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17240423#comment-17240423
]
ASF GitHub Bot commented on GEODE-8656:
---------------------------------------
kohlmu-pivotal commented on pull request #5775:
URL: https://github.com/apache/geode/pull/5775#issuecomment-735503985
I, like @onichols-pivotal do not see any problem with the original
[PR.(https://github.com/apache/geode/pull/5670). It was even signed off by
arguably one of the SMEs in this code area.
If there is some private, non-Geode, related failure, then is this really a
concern to Geode? ServerLocation is in the "internal" package space. Which
means that any Geode committers/developers have fullest right and privilege to
alter and modify its behavior without concern of breaking backward
compatibility.
@gesterzhou It seems there is a growing concern that this change has broken
a rolling upgrade and backward compatibility. I think there is a much larger
concern now. How did an internal class leak out of its "internal" containment?
How did we not catch THAT? WHY would an internal class now break rolling
upgrades?
I fully support the reverting of this PR if we can prove that rolling
upgrades will suffer from this change AND then raise a corresponding ticket to
fix the underlying root cause, which it seems is the leaking of internal
classes.
I disagree with the proposed fix of "adding of the memberId String to the
ServerLocation". Whilst it may "fix" the problem, it merely glosses over the
fact that we seem to have a potentially larger problem here.
@bschuchardt would you mind looking at this problem? As it seems that there
is a larger problem here that would possibly not have been considered when we
initially approved the [PR](https://github.com/apache/geode/pull/5670).
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
> Ping not sent to the right gateway receiver endpoint when several receivers
> have the same hostname-for-senders value
> --------------------------------------------------------------------------------------------------------------------
>
> Key: GEODE-8656
> URL: https://issues.apache.org/jira/browse/GEODE-8656
> Project: Geode
> Issue Type: Bug
> Components: wan
> Reporter: Alberto Gomez
> Assignee: Alberto Gomez
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.14.0
>
>
> When several gateway receivers have the same value for hostname-for-senders
> (for example when running Geode under kubernetes and a load balancer balances
> the load among the remote servers), the ping messages sent from the gateway
> sender client are only sent to the receiver (endpoint) to which the sender
> connected first.
> The other receivers will not get the ping message which will imply that the
> connections towards them will be closed after the configured timeout expires.
> This ticket is a continuation of the work done ticket GEODE-7565.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
