Storage on VMS

[Robert Brothers]

We get this message once in a great while on this system. Should I be concerned?

%%%%%%%%%%% OPCOM 7-DEC-2008 14:10:23.13 %%%%%%%%%%%
14:10:23.13 Multipath access to device $1$DGA407: has been auto switched from path PGD0.5000-1FE1-5002-EC6C (HECUBA) to path PGC0.50

%%%%%%%%%%% OPCOM 7-DEC-2008 14:10:24.11 %%%%%%%%%%%
14:10:24.11 Multipath access to device $1$DGA235: has been auto switched from path PGD0.5000-1FE1-5003-0F1C (HECUBA) to path PGC0.50

[Andy Bustamante]

You have some version of fibre channel storage. OpenVMS will switch paths if a path fails or is unavailable. Check your storage. If you need details on how to do this, post the version of server and storage. This could indicate a down path (controller, switch, cable) or someone has forced a path switch manually.

Andy Bustamante

[Robert Gezelter]

Robert,

Welcome to the ITRC OpenVMS Forum!

I heartily agree with Andy. Any error messages should be resolved. I have far too many crises where the situation has been brewing for a long time, but it was not yet fatal.

It could be nothing worse than a loose cable, but it is always worth resolving.

- Bob Gezelter, http://www.rlgsc.com

[Robert Brothers]

The wierd thing was it only failed a few of the luns not all of them.. I have 10 ES45's (4 hba's in each)and 4 EVA's no other errors any place that I could find. Could it move them for performance reasons?

[Uwe Zessin]

If both EVAs are running active/active (VCS V4) firmware, yes. active/passive EVAs do not 'optimize', because no I/O is possible through the paths presented by the controller which does not own the affected virtual disk(s).

[Hoff]

Check the error logs and check the EVA error logs.

It's (also) possible to manually toss disks around with SET PREFERRED_PATH, etc.

[Robert Atkinson]

Robert, I'm surprised nobody else has already said, but this is perfectly normal behaviour.

Both the EVA and VMS will switch disk paths automagically based on the load down any particular path or controller.

It's worth checking from the switches whether all connected ports are working OK, but these switches are part of normal operation, and generally occur just after boot.

Rob.

[Uwe Zessin]

From the WWNs those are quite old EVAs. An active/passive EVA will not automatically transfer virtual disk ownership between controllers.

Unfortunately, the OPCOM messages are truncated - we do not see to which EVA port the path has been switched to.

[Robert Atkinson]

Uwe, this is one of my 'boot' sequence path switches :-

%%%%%%%%%%% OPCOM 5-FEB-2009 04:48:23.90 %%%%%%%%%%%
04:48:21.23 Multipath access to device $1$DGA104: has been auto switched from pa
th PGA0.5000-1FE1-5006-118C (ALPHA) to path PGA0.5000-1FE1-5006-1188 (ALPHA)

I'm interested to know what part of the WWID gives away it's age, as these WWID's don't look that much different to Rob's?

Rob (A).

[Uwe Zessin]

The format is:

> 5000-1FE1-5006-1188

5 - IEEE Registered Name

000-1FE - IEEE OUI / vendor identification
"00-01-FE" is registered to DIGITAL EQUIPMENT CORPORATION

1-5006-1188 - vendor-specific code

Digital, Compaq & HP have been using this range for HSG80 + EVA GL/XL arrays.
The earliest WWN I have is from a HSG-80 training in October 1998 which starts with 5000-1FE1-0000-0C__
(last digits anonymized).

First EVA-5000: 5000-1FE1-0015-2D__ (January 2002)

and so on...

EVA-3000: 5000-1FE1-5001-A5__ (sometimes in-between they was a jump to '5000-1FE1-500.' which includes the EVA-5000).


About your configuration:
Assuming that nobody has messed with the controller names: VMS has switched from Controller-B, Port_1 to Controller-A, Port_1.

My guess is that this is an active/active EVA and you are running a recent version of OpenVMS which understands optimized/ non-optimized paths. Take a look in Command View EVA whether the virtual disk that represents $1$DGA104: is owned by Controller-A (vdisk properties - presentation tab).


Before somebody asks: I am not aware that the array model is encoded in the WWN.