ESX hosts lose SAN paths after switch firmware upgrade

[johnCatBE]

Hello

We're in the process of upgrading SAN switch firmware (5.3.0x to 6.0.0c) in a dual-fabric mixed environment (ESX Polyserve EVA CA Sun cluster )

Did Fabric A switches last evening .. no probs with unix hosts or even Polyserve cluster members but we lost VMware hosts - ie no failover to other untouched fabric.

Details are ESX 3.5 / UP4
HP Blades BL465c (G) using emulex lightpulse cards.

Has anyone experienced similar issues? I was concerned that it might be due to compatibility with 6.0.0c firmware?

thanks

[Rob Buxton]

Once lost some guests due to some time-out issues. There are some time out values you can set. But otherwise we've successfully updated SAN Switch firmware.
Were the LUNs showing correctly on both HBAs through the VC Console?

[johnCatBE]

Yes our current theory is that VCB backups were underway on the impacted ESX servers at the time of the issues...........and that VCB might not cope with even a temporary hiatus in its storage............so in the event of nothing else coming to light were going to suspend or avoid our VCB backup windows when doing SAN maintenance.


Thanks

John

[Rob Buxton]

Also, are you stopping at 6.0.0c? There are further updates. 5 to 6 is quite a jump, but that shouldn't have caused issues. You may want to look at updating to the latest release.

[johnCatBE]

rob

No...........we're off to 6.1.1d in a couple of weeks time..........wanted to try and understand the ESX/VCB issue first though.

john

[Uwe Zessin]

Yes, VCB is very unforgiving as it seems to do a lot of SCSI reservations.

Rob, are you talking about the Windows timeout values?

from cmd:
> reg query "HKLM\SYSTEM\CurrentControlSet\Services\Disk" /v TimeOutValue
> reg add "HKLM\SYSTEM\CurrentControlSet\Services\Disk" /v TimeOutValue /t REG_DWORD /d 60

[Rob Buxton]

Uwe,
No there are timeout settings for ESX, Solaris etc. that go into the config files. We had some issues where we experienced timeouts with a link loop failure on the EVA. Windows servers typically survied but we had issues on ESX and Solaris Server. Our local HP guy gave us some values to add. Not at work so I don't have access to the parameters that were updated.

[Niels Vejrup Pedersen]

Hi,

I've updated a lot of switches from 5.3 to 6.1.2 and 6.2.1 lately.

Doing Fabric A first and Fabric B afterwards without any issues.

To my knowledge the default timeout setting is 60 secs on VMware ESX 2.5 and onwards.

Regards