I am seeing path updation issues with multipath tools versions 0.4.4 and 0.4.5, when tested with EVA HSV 200, MSA and XP (Active Active arrays) on RHEL U1.
Failover works fine and fails to the other path in the path group, but when the failed path comes back online, the status of this path is [ready][failed]. This doesn’t change unless “multipath” is executed manually from the command prompt.
Some of the unique observations seen are:
In MSA, when failed path comes back online, this path and currently active path goes to [ready][failed] state.
In EVA HSV 200, where the path groups are created using “group_by_serial” policy,
If IO is pumped on a dm device, a path is failed and when the failed path in a particular path group (each path group having 2 paths to 2 ports of the controller) comes back online, the status on these paths in that path group is going to [ready][failed], then the IO starts continuing on the other path group.
Any pointers for this would be really helpful. I would like to know if this kind of behaviour (partially or fully) is seen in any other storage systems?
In general when a path can go to [ready][failed] state?
Thanks and regards