topic Re: NSR - Identifying polling hosts in StoreEver Tape Storage

NSR - Identifying polling hosts

smw — Fri, 06 Nov 2009 13:17:12 GMT

We have several servers (primarily Red Hat, but on a couple of occasions HPUX) which cannot see all our tape drives. They can see all of them fine when the servers are first booted up, and can be used perfectly. After a short while (usually a few days) they stop being able to use them/see them via DP.

An HP engineer who was working on another case for us, suggested that he had seen problems before when Windows machines poll SAN devices excessively, causing problems for other systems.

So viewing NSR trace logs, I see the following snippits, which I believe is one of these machines polling:

0ms 27us PS_DrvSend: Bus:1 Target:2 Lun:0 CDB:x00
0ms 0us SCSI msg out c0000000 cdb 00000000000000000000000000000000
0ms 378us FrmHdr 06040100 00041a00 08290000 06000000 01a0ffff Port 0
0ms 2us FCP_LUN 00010000 00000000 FCP_CNTRL 00000000 FCP_DL 00000000
0ms 0us FCP_CDB 00000000 00000000 00000000 00000000 IOCB 80F3397C
0ms 18us RMI_getPortIdforRoute: routeIndex: x4, Route_Port: x0
0ms 3us fcpTrns_cleanupPersistentCommand: Leaving ....

So my question is, how can one identify which host/s are doing this? They seem to be doing it several times a minute.

Failing this, how can one fully rescan devices within Red Hat, as it does on start up? An hp_rescan has only once managed to reclaim any devices, yet a reboot works each time, but obviously is not desirable.

All our firmware and drivers are current.

Thanks for any help.

Re: NSR - Identifying polling hosts

Marino Meloni_1 — Mon, 09 Nov 2009 09:05:02 GMT

Can you attach the nsr report in *.mth format?

Re: NSR - Identifying polling hosts

smw — Mon, 09 Nov 2009 16:14:06 GMT

I've attached our NSR report. Thank you for your help.

Re: NSR - Identifying polling hosts

Marino Meloni_1 — Mon, 09 Nov 2009 16:38:03 GMT

seems you have polling going on,
I hope you can identify it with the attached decoded nsr report

Re: NSR - Identifying polling hosts

smw — Tue, 10 Nov 2009 11:56:23 GMT

Thank you for your help. Hopefully we can have a look at what is polling.

Strange thing is we did a DR-test this week and even though none of the problem-servers were failed over or rebooted etc, they can now see all the tape drives. Some of the servers are in the same room as the tape drives they couldn't see, yet by failing the OTHER room over + back again, they now work. It'll be interesting to see if this continues.

Anyway, thank you for your assistance.

If anyone else knows if there is a good alternative to hp_rescan on Red Hat, please reply.