Red Hat Bugzilla – Bug 1025456
mirror device failure scenarios lead to deadlocks (part 2)
Last modified: 2016-04-08 17:19:50 EDT
The test listed in the last comment seems to fail one device first and then a second. The second failure happens while the mirror is being up-converted due to the first repair.
I need to know a couple things to determine whether to fix this issue:
1) is it reproducible (It'd be useful to see those results and compare)
2) what is the result of the failure? Is the mirror usable at that point or not? Failure at an intermediate stage would be the same as bug 1016296, and I would likely close WONTFIX if there is no solution. (RAID does not suffer from these issues.)
It is possible that we just need to cope with the fact that the final repair is called on the up-converting tmp image. This could be a simple fix but depends on the answer to #2.
Moving to rhel6.7 for further consideration. Whether to close or not will depend on the answers to the questions in comment 1.
(In reply to Jonathan Earl Brassow from comment #2)
> Moving to rhel6.7 for further consideration. Whether to close or not will
> depend on the answers to the questions in comment 1.
You are welcome to reopen this bug to be reconsidered if answers are provided to comment 1.
Filed bug 1325457 for the specific multiple asynchronous mirror device failures mentioned in that last few comments of this bug.