Red Hat Bugzilla – Bug 479330
cluster crashes without mcast-max workaround
Last modified: 2015-11-15 19:06:32 EST
Description of problem:
If --cluster-mcasts-max is set 0, the cluster crashes with corrupted frames.
Version-Release number of selected component (if applicable):
How reproducible: easy
Steps to Reproduce:
[aconway@mrg10 tests]$ ../qpidd --cluster-read-max 0 --cluster-mcast-max 0 --load-module ../.libs/cluster.so&
[aconway@mrg10 tests]$ ./perftest --count 10000 --nsubs 4 --npubs 4
Cluster crashes with log similar to:
2009-jan-08 16:14:46 critical 22.214.171.124:27404(READY) error in cluster delivery: framing-error: Framing version unsupported (qpid/framing/AMQFrame.cpp:80)
2009-jan-08 16:14:46 notice 126.96.36.199:27404(LEFT) leaving cluster aconway
This is sometimes, but not always accompanied by an openais crash.
The mcast-max and read-max limit the number of outstanding multicasts to openais, so this appears to be a flow control problem. Looks related to:
but there may be a qpidd bug lurking as well.
Resolved by openais patch ipc-14. The cluster-mcast-max option is removed.
Validated on RHEL 4.7 / 5.2 i386 / x86_64 that --cluster-mcast-max 0
is not able to specify and therefore not able to get framing-error: Framing version unsupported issue.
Framing-error issues drastically reduced by new openais version 0.80.5-2. No such issue found after releasing that new openais version.
An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on therefore solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.