SCSI Sense commands

I just can't figure this one out.
The ESX hosts we have in the environment are reporting issues with connectivity to the storage fabric.  They connect fine and push data back and forth but one server in particular reports that it keeps losing connectivity to one of the paths through powerpath.  Then it fixes itself.

With the vkernel log it shows SCSI Sense Codes and points to Destination then the Host and back and forth.   When I get the time this has to be resolved.


Mike D said…
We have the same problem here - Win 2008 R2 VM with the PVSCSI driver, vsphere 4.1, Powerpath VE, Qlogic CNAs, over an FCoE fabric. We've had EMC, Cisco, Qlogic, and VMWare looking at it but no one has come up with anything.
virtualvelcro said…
We ended up checking the cabling between the ESX host and the fiber switch which fixed some of the errors. It came to a head a couple weeks ago when we had one LUN disappear which crashed 5 VMs, three Hosts disconnected and wouldn't reconnect even though hostd and vxpa were running fine. (And it wouldn't even connect directly with the viclient)

We had to reboot the Storage Processors, after the process the ESX that had the lock on the LUN PSOD. This cleared the lock and I haven't seen much for errors on the servers.

EMC pointed this article as part of the post-mortem.

They gave the issues at the time of the support call as a Powerpath issue. Being you are using 4.1 have you upgraded your powerpath/ve to the newest rev?
Mike D said…
Yeah, we upgraded to the latest PP/VE. We were thinking this could actually be something happening at the guest level. We were not able to reproduce the error on demand, but then we vmotioned a host and the error occurred at the same time. We ultimately shut that host down and haven't seen the error since that host has been down. How can a host cause storage paths to drop? Dunno...