B.10.6. RHBA-2011:0854 — corosync bug fix update

When the corosync server terminated unexpectedly, if it was connected to corosync clients, a shared memory leak occurred. This bug has been fixed and no memory leaks occur in the described scenario.

When a ring ID file was smaller than 8 bytes, the corosync server terminated unexpectedly. With this update, if no proper ring ID file can be loaded, the corosync server creates one and no crash will occur.

During the recovery phase, the corosync server sometimes terminated unexpectedly. As a consequence, a network token was lost and new configuration had to be created. This bug has been fixed and the corosync server no longer crashes in the described scenario.

In rare circumstances involving multiple running nodes, the corosync server terminated unexpectedly during shut down. This bug has been fixed and the corosync server no longer crashes.

When inconsistent cluster.conf files with different versions were used among nodes, a memory leak occurred in the corosync server during the configuration reload. This bug has been fixed and the configuration reload via the cman_tool no longer causes memory leaks.