VSphere 5.0 "Lost connectivity to storage device" even though it's detached

[bsti]

Hello,

  I have a VSphere 5.0 cluster with several hosts.  I have an operation that runs every few hours that detaches and reattaches a set of RDM LUNs to some of the VMs in the cluster.  I've followed the "how to unpresent a lun" procedure to the letter, however I still see the "lost connectivity to storage device" error message in the logs.  The VMKernel logs have even uglier error messages, indicating it's lost paths to storage unexpectely.  I believe this activity causes my hosts to become unresponsive from time to time.

Here is the procedure I use to dismount an RDM LUN:

1)  Set the disk offline in the guest OS

2)  Remove  the disk from the VM and delete the mapping file

3)  Detach the RDM lun from ALL ESXi hosts

4)  Unmap the LUN on the storage array so the hosts no longer see it

5)  Rescan all hosts

These are the relevant log messages I see:

Device naa.60a980003753314d733f434b2f2f4e4e, has been turned off administratively.
XXX esx.problem.scsi.device.state.off.category not found XXX
3/8/2013 1:34:33 PM
host2.mydomain.com

Task: Detach SCSI LUN
info
3/8/2013 1:34:34 PM
Detach SCSI LUN
host2.mydomain.com

Lost connectivity to storage device naa.60a980003753314d733f434b2f2f4e4e. Path vmhba3:C0:T0:L22 is down. Affected datastores: Unknown.
error
3/8/2013 1:35:56 PM
host2.mydomain.com

Lost connectivity to storage device naa.60a980003753314d733f434b2f2f4e4e. Path vmhba2:C0:T0:L22 is down. Affected datastores: Unknown.
error
3/8/2013 1:35:56 PM
host2.mydomain.com

Lost connectivity to storage device naa.60a980003753314d733f434b2f2f4e4e. Path vmhba3:C0:T1:L22 is down. Affected datastores: Unknown.
error
3/8/2013 1:35:56 PM
host2.mydomain.com

Task: Rescan all HBAs
info
3/8/2013 1:35:59 PM
Rescan all HBAs
host2.mydomain.com

Task: Rescan VMFS
info
3/8/2013 1:36:02 PM
Rescan VMFS
host2.mydomain.com

My specific questions are:

-  Is this normal behavior?  This seems like something is not going correctly to me, but I could be wrong.

-  If not, is there anything wrong with my procedure above?

Thanks for the assistance!

[mhost]

Hello,

Our experience is the same - although for a VMFS datastore.

We have just removed a datastore, following the same procedure

• Unmount datastore on each host
• Detach storage device on each host
• Unpresent LUN from SAN
• Rescan for storage on each host

And we still received the "Lost connectivity" alert when unpresenting from SAN.

I have posted my question here, before i saw yours.

Best regards

Martin

[SG1234]

between 2) and 3) could you do -- esxcfg-scsidevs -o <naaid> -- naa id of the lun can be obtained from esxcfg-mpath -l

hope this helps

~Sai Garimella

[bsti]

Hi!  Thanks for  the reply.  I'd love to see the discussion you cited but the link you gave doesn't work.  I'm very interested in any details about this.  IT's entirely possible this is a non-issue, but I occasionally get errors on my hosts where they mysteriously lose connectivity with VCenter and I wonder if its somehow related.

Thanks!

[bsti]

Thanks for the suggestion.  I will likely give this a try.

[bsti]

I tried using the esxcfg-scsidevs -o command today. It had no effect.  In fact, I can't even tell what effect it had at all.  What is this command supposed to do?

THanks for the suggestion.  Any others?

[mhost]

Jeg afholder ferie i perioden 25/3 t.o.m. 2/4.

Mails besvares hurtigst muligt efter min tilbagevenden.

Du kan alternativt kontakte Servicedesk på: servicedesk@vejle.dk eller tlf. 7681 1111

Med venlig hilsen

Martin Holst | IT-konsulent | IT-Drift og Support

Vejle Kommune | Skolegade 1 | DK-7100 Vejle

mhost@vejle.dk<mailto:mhost@vejle.dk> | * 45 76811117 | È45 24200054

P tænk på miljøet før du printer denne e-mail