Bug 872696 - QMF queue events do not propagate to HA backups
Summary: QMF queue events do not propagate to HA backups
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: None
Product: Red Hat Enterprise MRG
Classification: Red Hat
Component: qpid-cpp
Version: Development
Hardware: Unspecified
OS: Unspecified
high
high
Target Milestone: 2.3
: ---
Assignee: Alan Conway
QA Contact: MRG Quality Engineering
URL:
Whiteboard:
Depends On:
Blocks: 698367
TreeView+ depends on / blocked
 
Reported: 2012-11-02 18:45 UTC by Jason Dillaman
Modified: 2013-03-19 16:39 UTC (History)
5 users (show)

Fixed In Version: qpid-cpp-0.18-9
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2013-03-19 16:39:08 UTC
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)
Quick patch to prevent duplicate output tasks (1.81 KB, patch)
2012-11-08 19:47 UTC, Jason Dillaman
no flags Details | Diff
Quick patch to prevent duplicate output tasks (2.07 KB, patch)
2012-11-08 21:11 UTC, Jason Dillaman
no flags Details | Diff


Links
System ID Private Priority Status Summary Last Updated
Apache JIRA QPID-4430 0 None None None 2012-11-12 21:26:13 UTC

Description Jason Dillaman 2012-11-02 18:45:43 UTC
Description of problem:
After unsubscribing the last consumer from an auto-delete queue, the delete event was not propagated to the backup nodes.  Additionally, after re-creating a queue with the same name, the the backup did not receive the declare event.  The original queue declare event was correctly propagated to the backup.

Version-Release number of selected component (if applicable):
qpid-cpp-0.18-6

How reproducible:
100%

Steps to Reproduce:
1. Subscribe to an auto-delete queue
2. Unsubscribe from the auto-delete queue
  
Actual results:
The backup node did not receive the delete event.

Expected results:
The backup node should receive the delete event.

Comment 1 Jason Dillaman 2012-11-05 15:48:53 UTC
After additional investigation, the issue appears to have been exposed due to channel errors that resulted in the link reconnecting (see BZ873347).  The QMF channel did not have any errors in the log but somehow the backup missed a queue creation event and thus failed to replicate a queue.

Comment 2 Jason Dillaman 2012-11-07 17:55:55 UTC
Issue is not 100% reproducible, but has happened often across numerous machines.

Comment 3 Jason Dillaman 2012-11-08 19:47:04 UTC
Created attachment 641055 [details]
Quick patch to prevent duplicate output tasks

Comment 4 Jason Dillaman 2012-11-08 19:49:00 UTC
Discovered that the inter-broker link had hundreds of thousands of enqueued OutputTasks -- representing only a few unique consumers.  There should only be only a single output task for a given consumer.  This appears to have stalled the delivery of QMF messages to the backup broker.

Comment 5 Jason Dillaman 2012-11-08 21:11:33 UTC
Created attachment 641080 [details]
Quick patch to prevent duplicate output tasks


Note You need to log in before you can comment on or make changes to this bug.