Hi all, First off, keep up the good work!
I have a small cluster running Pacemaker 1.0.9 and OpenAIS/Corosync, as checked out from the Debian lenny-backports repository (amd-64). I think I may have come across this bug: https://lists.linux-foundation.org/pipermail/openais/2010-June/014848.html https://lists.linux-foundation.org/pipermail/openais/2010-February/013812.html I would get three instances of corosync at startup, and nodes would sometimes have issues coming online from the view of the cluster (they could communicate, but would register as OFFLINE). I'm using rsyslog, at runlevel S10. From what I can tell, out-of-the-box corosync is started by (Debian-specific?) rcS.d instead of a numbered rc?.d like rsyslog. Despite using a numerically-higher runlevel than rsyslog, this discrepancy seemed to cause the above issues. I was able to resolve the problem by removing all rc.d entries for corosync and re-instating them as the default S20. I am left with two questions: -Is there an immediately-obvious reason why rcS.d should be used instead of the defaults? i.e. Can I expect future problems with my "fix"? -If this isn't the place to report this (perhaps if it's distro-specific) to whom should I report it? Thanks! -Arthur
_______________________________________________ Openais mailing list Openais@lists.linux-foundation.org https://lists.linux-foundation.org/mailman/listinfo/openais
