VMware Cloud Community
KWKirchner
Enthusiast
Enthusiast

SQL Server DB Disk become unreadble after reboot - NOT offline

System:

ESXi v6.0 on HP DL380

Microsoft SQL 2014

NetApp AFF8040 SAN running OnTap 9.1p1

Windows Server 2012R2

We have seen this occur 3 different times at 3 different sites. The SQL server is running fine, gets cleanly shutdown, and upon bootup the VMDK dedicated to the DB (.mdf files) is off-line and uninitialized. The only course of action is to reformat the drive and restore from a DB backup.

There are 8 disks/VMDK's on these servers and it is always the database vmdk that takes it in the shorts. The other 7 drives are fine.

I would not be completely surprised (except for the clean shutdown) if just the DB was corrupted, but for the entire filesystem on the drive to be rendered useless, that has me at a loss.

Does anyone have any idea what might be happening?  Any recommended tools to repair a disk/vmdk that apparently has deep filesystem issues?

0 Kudos
3 Replies
hussainbte
Expert
Expert

Are you using vmdks or RDMs.?

when the DB boots up,,you have any other option other than formatting the dives..? like online the disk.. I am trying to check if they are only going offline pr losing the partition information.

if you can Also share the vmkernel and vmware.log of the vm whan the vm boots up..

If you found my answers useful please consider marking them as Correct OR Helpful Regards, Hussain https://virtualcubes.wordpress.com/
0 Kudos
KWKirchner
Enthusiast
Enthusiast

We had some severe cabling issues between our NAS and the switch. I cant conclusively put the blame there because the switch is smart enough to deactivate any port-channel members that are not working. That said, we only had 2 out of the 12 10Gb connections actually working.  I would expect that to possibly cause a performance issue, but not corruption.

We have since fixed all of the port-channel issues and have our fingers crossed that we wont see the issue again.

To your questions:

     The VM boots up just fine. It had 8 VMDK's attached and all but one of those came up properly. The VMDK that stored the actual DB files was seen by the Windows 2012R2 OS as being there, but it apparently had no NTFS or partition structure that it recognized. Brining the disk on-line didnt have any affect on the drive readability.  The partition info was gone.

0 Kudos
jorgegabriel
Contributor
Contributor

I know this is an old post, but have you found a solution or root cause of this issue, I'm experiencing exactly the same but in vSphere and ESXi 6.5 

0 Kudos