Dead path with IBM ds3500

[alicem]

Our setup: two IBM servers each with two SAS HBA connected directly to a DS3500 (DS3512) SAS storage.

The problem is that after some I/O operation one of the paths to storage is lost (Status: dead). However it is restored after HOST reboot. This problem occurs on both of the hosts. (not simultaneously)

After the path lost the Controllers on the storage continuously change the ownership of LUNS.

We face this problem with ESXi4.0 after that we upgrade the hosts to ESXi5.0 (469512) but face the problem again.

The paths are defined as MRU as advised by VMware.

Any advice will be highly appreciated

Regards,

[prince1121]

Hi

please change to this parameter results in LifeKeeper waiting  approximately  80 seconds before determining that an unresponsive system  is dead, rather than the default wait time of approximately 15 seconds.

____________
Best of luck

[alicem]

prince1121,

the symptom is; one of the path to storage system becomes "dead" not the host system. (and in some cases virtual servers are inaccessable)

(Though I agree on incresing Lifekeeper parameter to some higher value...)

[alicem]

TMryan,

Our system seems OK.

As I say I am not sure if it is firmware upgrade,VT-d disable or the wrokaround from vmware.

My advise is to disable VT-d if you dont need it.

If the issue cont. then please post storage and/or esx logs so that I/we can comment more.... (may  be a new post will be better...)

Regards,

Ali Cem

[fum-hba]

Hi,

we ran into the same issue.

First I did the steps recommended in VMware KB 1030265, but the Problem still exists.

Now disabled VT-d. Maybe the problem will be fixed.

The IBM Hotline knows about the problem, but wont help because its not supported.

[alicem]

Hope it solves the issue fum-hba,

How could this kind of an industrial standard ( DS3500, xSeries Servers, SAS connection) environment not supported/tested by IBM…

I simply could not make head or tail if this…

Refusing support for this kind of a standard enviroment is simply unaccaptable in my point of view....

Regards,

[fum-hba]

Hi,

since I disabled VT-d in the BIOS/UEFI settings, the problem is solved. Two installations run without any problems.

[infusednz]

I've just run in to this now with a brand new build. It was sold to us as a working solution.

Is the DS3524 with SAS now supported under VMWare 5?

[darioj]

As far as I remember, SAS has been on the IBM supported list since ESX 3.x.

We have SAS attached xServers to a ds3500, never had such problems.Started with ESXi 4.1, upgraded up to 5.0 U1, U2.

Always check the IBM SSIC and Server Proven, update the hardware firmware ( servers and san ) before ESXi.

Never upgrade to the latest VMware relaease, wait for the first service pack

Hope this helps

Cheers

Dario