In our HA cluster consisting of two nodes (versions: openais-0.80.3-26.8.1, pacemaker-1.0.5-0.5.6, libdlm-2.99.08-8.11.11), we noticed the following behavior concerning access to files residing on our shared ocfs2 file system managed by OCF resource agent "Filesystem":
- stopping openais on designated coordinator, it is possible to access shared file system after DC has changed; - reset (init 6) on designated coordinator, it is also possible to access shared file system after DC has changed; but: immediate power off on designated coordinator, it is not possible to access shared file system after DC has changed; it seems some locks on shared file system prevent accessing file system from new DC !!! Having a look on concerned processes by ps ax, we see process state D (Uninterruptible sleep (usually IO)) !!! Any idea to solve the problem ??? Regards, Armin Haussecker
_______________________________________________ Openais mailing list [email protected] https://lists.linux-foundation.org/mailman/listinfo/openais
