HPE EVA Storage

MSA 2324i - RAW Partition recovery to NTFS

 
Adrian Barker_1
Occasional Advisor

MSA 2324i - RAW Partition recovery to NTFS

Hi,

We have a problem with our MSA which will occasionally result in an NTFS volume being picked up as RAW by Windows (2008 R2 Hyper-V Cluster). A snapshot taken after the event can be mounted as NTFS without any issues. Rolling the master volume 'forward' to the known working snapshot results in a RAW volume again.

I've tried chkdsk and the usual recovery suspects but it's as if the drive is read-only or something and I can't resurrect the NTFS file system. I've also tried changing the LUN incase that would help Windows see it as a different disk. We also get the same behaviour if accessing the SAN from a fresh Windows installation which I think rules out Windows as the source of the problem.

My problem is that if we have this problem for real (we already have) I have several hours of taking the data off the snapshot in order to take even longer putting it back on a newly created volume.

The background to this problem incase it helps was: In the event of a switch failure, the MPIO recovery time was too long, around 2 minutes. After playing around we've got this down to 30 seconds or so. While doing this we encountered several instances where, on recovery, one or more of the volumes is seen as RAW rather than NTFS in Windows. This has happened before the MPIO settings were changed.
2 REPLIES 2
Adrian Barker_1
Occasional Advisor

Re: MSA 2324i - RAW Partition recovery to NTFS

OK, a bit more information: I can mount the volume OK if I set the access mode to read-only. I need to restore read-write access though.

Looking at WireShark; there is a SCSI Reservation Conflict when accessing the LUN with write access. So, I gues my question now is how do I remove a persistent scsi reservation on the MSA?
Adrian Barker_1
Occasional Advisor

Re: MSA 2324i - RAW Partition recovery to NTFS

Wow, the danger of using technology that you don't know inside out;

The persistent reservation is put on by the Windows cluster services (obvious really) and can be forcibly removed by using the cluster.exe command;

cluster node /clearpr:[disk_id from disk management console]

Only use this command if you are sure that nothing else on the SAN is genuinely using the LUN. Hope this helps someone else!