Veritas-bu

[Veritas-bu] HP LTO3 FC drives "locking up" [C1]

2007-03-06 20:38:18
Subject: [Veritas-bu] HP LTO3 FC drives "locking up" [C1]
From: ewilts at ewilts.org (Ed Wilts)
Date: Tue, 06 Mar 2007 19:38:18 -0600
On 3/6/2007 3:09 PM, misha.pavlov at sgcib.com wrote:
> Drives "lock" up and becomes iresponsive.
> Front panel light show no signs of problem with solid green light on.
> Pressing eject button does not do anything.
> Brocade 4100 switch port does not "see" the drive anymore and shows 
> "In_Sync" instead of the "Online".

Depending on where you look, the drive can appear to the switch to be 
both In_Sync and Online.  A portshow will show both on functioning 
drives.  A switchshow will just show Online.

If the drive really has disconnected from the fabric then you need to 
figure out why.  Don't bother looking at NetBackup - there's nothing 
NetBackup can do if the drive has left the fabric.

For starters, clear the errors on the switch port.  Then the next time 
it happens, check the port stats - grab a screen shot of all of them. 
Look at the them during normal operation as well to see what the pattern is.

> The only way to bring the drive back is the powercycle.

Instead of doing a powercycle, see if you can do a portdisable followed 
by a portenable.  That will force the drive to attempt to log back in to 
the fabric again.

Also make sure the port speed is configured - don't let it autosense - 
and check the switch logs.  Make sure the fibre is clean and that you 
haven't exceeded the fibre distance.  Make sure you don't have any tight 
loops in your fibre path.

> A minute or two after the powercycling I can eject the tape and see
> fctl: [ID 517869 kern.warning] WARNING: 3799=>fp(1)::N_x Port with 
> D_ID=150500, PWWN=500104f0005ddda2 reappeared in fabric
> in /var/adm/messages
> 
> Drives and library are at the latest f/w revision.
> SUN and STK are clueless, but still looking for the last week.

Have you logged a call against the Brocade switch?  Have you made sure 
that the switch firmware is also current?  Do you know how to interpret 
the switch logs?

        .../Ed

-- 
Ed Wilts, Mounds View, MN, USA
mailto:ewilts at ewilts.org