Description of problem:
# cman_tool join
Relax-NG parser error : Reference SERVICE has no matching definition
Version-Release number of selected component (if applicable):
Steps to Reproduce (cluster-wide):
1. run while true; do cman_tool join; cman_tool leave; done
2. feel lucky
3. hit ctrl-c couple of times
it may happen that the regenerated schema gets corrupted and cluster operations on that node are disabled until the error is resolved.
* nothing like corruption happening
* message telling users about ccs_update_schema
* removing the cache automatically
Created attachment 520048 [details]
The fix simply invalidates the ccs_update_schema cache if any trap is received, forcing a schema regeneration in the next run.
I am unable to reproduce the issue reported after applying this patch.
Jaroslav, I left the manually patched version of ccs_update_schema on the marathon cluster (all 5 nodes), for you to stress some more, and should be able to provide a build by Monday.
Works well with the patched version, I'm not able to reproduce it any more.
The easiest way how to reproduce this is start the cman_join cman_leave loop on all nodes and kill -9 corosync on one, wait for messages like "cman_tool: Error leaving cluster: Operation already in progress" appearing rapidly and then try to stop the loop (ctrl-c, ctrl-q). Eventually it won't be able to start again.
Thanks Fabio, I'll wait for an official build.
can't reproduce it in any longer with cman-220.127.116.11-19.el6.x86_64
Technical note added. If any revisions are required, please edit the "Technical Notes" field
accordingly. All revisions will be proofread by the Engineering Content Services team.
Do not document.
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.