I am pleased to announce the release of Corosync 1.2.1. This update operates more smoothly with Pacemaker, especially in the shutdown case. This bugfix update has the following changes:
* Fix problem where loss-prone networks can generate false positive node failure detection. * Fix lockup which can occur during shutdown because of pthread use inside atexit handlers. * All thread-unsafe posix api usage was changed to use the thread-safe variants. * Fix race condition related to cpg_join/cpg_finalize/cpg_leave in a loop. * Fix error handling on service connects to not leak memory if the connection could not be made. * Fix token hold cancel's parameter which was causing Wireshark to coredump. * Major rework of the shutdown process to be more friendly with service engines, specifically Pacemaker. * Remove a segfault in the cfg api. * Resolve a problem where if a node fails and is restarted within the token timeout parameter, it wasn't being detected as a failure. It is now detected faulty first then joins back into the network. * Remove invalid assertion in totemsrp which could trigger an assertion. * init script rework to fix problems found during Pacemaker integration * Resolve problem where if consensus = 1200 (default) corosync would print an error and not start. A big thanks to our community who investigated and resolved these issues. The software is available via our website at: http://www.corosync.org. Regards -steve _______________________________________________ Openais mailing list [email protected] https://lists.linux-foundation.org/mailman/listinfo/openais
