[
https://issues.apache.org/jira/browse/GEODE-8333?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17188044#comment-17188044
]
ASF subversion and git services commented on GEODE-8333:
--------------------------------------------------------
Commit 27b8e47e9f6e433dc71bffc5420ebcb51f79797e in geode's branch
refs/heads/develop from Jens Deppe
[ https://gitbox.apache.org/repos/asf?p=geode.git;h=27b8e47 ]
GEODE-8333: Change Redis adapter threading model - fixes pubsub issues (#5488)
- Do not use Netty threads for the entire request lifecycle. Each
instance of ExecutionHandlerContext (essentially each client
connection) uses a command queue which is actioned by a single thread
taken from the Geode 'waiting pool'.
- Every SUBSCRIBEed client is moved to a separate EventGroupLoop so that
PUBLISHed messages are not sent back using the 'normal' Worker
EventLoopGroup. This avoids a deadlock issue where PUBLISHed messages
need to be sent using the same thread that the PUBLISH response needs
to happen on.
- Fix issues with PubSub where switching the EventLoopGroup may fail
(because a client has already closed the connection) thus resulting in
hanging PUBLISHers waiting for a Subscription to be marked as
'readyForPublish'.
- Consolidate various MockSubscriber classes
> Fix PUBSUB hang
> ---------------
>
> Key: GEODE-8333
> URL: https://issues.apache.org/jira/browse/GEODE-8333
> Project: Geode
> Issue Type: Bug
> Components: redis
> Reporter: Sarah Abbey
> Assignee: Darrel Schneider
> Priority: Major
> Labels: pull-request-available
>
> PUBSUB hangs with concurrent publishers and subscribers on multiple servers.
> The initial fix is being reverted because it caused a bug in which responses
> to clients could now be out of order.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
