Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Hi all! We have been experiencing serious problems with heartbeat recently, although the problem may not be related with heartbeat directly. When trying to stop heartbeat on the first node and waiting for the complementary node to take resources over, the directory of WEB resource cannot be unmounted and the following messages can be found in heartbeat's log: Filesystem[5094][5154]: 2006/08/20_21:32:55 INFO: No processes on /storage/web were signalled Filesystem[5094][5154]: 2006/08/20_21:32:55 INFO: No processes on /storage/web were signalled Filesystem[5094][5157]: 2006/08/20_21:32:56 ERROR: Couldn't unmount /storage/web, giving up! Filesystem[5094][5157]: 2006/08/20_21:32:56 ERROR: Couldn't unmount /storage/web, giving up! Filesystem[5029][5159]: 2006/08/20_21:32:56 ERROR: Filesystem Generic error Filesystem[5029][5159]: 2006/08/20_21:32:56 ERROR: Filesystem Generic error Then heartbeat reboots the machine after declaring itself dead: ResourceManager[3186][5259]: 2006/08/20_21:32:56 CRIT: Resource STOP failure. Reboot required! ResourceManager[3186][5259]: 2006/08/20_21:32:56 CRIT: Resource STOP failure. Reboot required! ResourceManager[3186][5260]: 2006/08/20_21:32:56 CRIT: Killing heartbeat ungracefully! ResourceManager[3186][5260]: 2006/08/20_21:32:56 CRIT: Killing heartbeat ungracefully! I've also tried to unmount /storage/web via "umount" command interactively, however it's still saying it's busy :-( The cluster of 2 nodes has been running fine for over 9 months without any serious problems. This problem started to happen recently (about 2 weeks ago) with heartbeat version 2.0.1 (which has been in production use for the mentioned 9 months), and the upgrade to the latest heartbeat version, 2.0.7, didn't improve situation. Can anybody please tell me how to better figure out why it can't be unmounted? /storage/web is a mounted DRBD device, using ext3 filesystem on top. Thank you for you answers! Mirek