ESX 3.5u5
While committing snapshots during an image backup operation we had problems with a VM that died and could not be started again until we rebooted the ESX host. The logs show that there were problems with the journal:
Feb 21 03:08:26 trvmp02 vmkernel: 31:09:31:32.441 cpu9:1211)WARNING: FS3: 2460: Local host has lost lock [type 10c00001 offset 79489024 v 2977, hb offset 3809280
Feb 21 03:08:32 trvmp02 vmkernel: 31:09:31:38.324 cpu2:1045)WARNING: J3: 1688: Failed to create new txn because journal is unavailable
Feb 21 03:08:32 trvmp02 vmkernel: 31:09:31:38.324 cpu2:1045)WARNING: J3: 1688: Failed to create new txn because journal is unavailable
Feb 21 03:09:16 trvmp02 vmkernel: 31:09:32:21.893 cpu12:1261)WARNING: SCSI: 119: Failing I/O due to too many reservation conflicts
Feb 21 03:09:16 trvmp02 vmkernel: 31:09:32:21.893 cpu12:1261)WARNING: FS3: 2429: Couldn't verify lost lock: SCSI reservation conflict
Feb 21 03:09:16 trvmp02 vmkernel: 31:09:32:22.026 cpu0:1129)WARNING: SCSI: 119: Failing I/O due to too many reservation conflicts
Feb 21 03:09:16 trvmp02 vmkernel: 31:09:32:22.026 cpu0:1129)WARNING: FS3: 2429: Couldn't verify lost lock: SCSI reservation conflict
Feb 21 03:09:21 trvmp02 vmkernel: 31:09:32:27.008 cpu1:1129)WARNING: SCSI: 119: Failing I/O due to too many reservation conflicts
Feb 21 03:09:21 trvmp02 vmkernel: 31:09:32:27.008 cpu1:1129)WARNING: FS3: 2429: Couldn't verify lost lock: SCSI reservation conflict
The only reference to such an error that I've found is this kb article that was written for ESX 3.0.x: http://kb.vmware.com/kb/1003925
It doesn't have any advices for how to prevent this. This has now happened twice in this environment and I wonder if anyone has seen something similar and know how we can avoid this from happening.
Lars
