PERHAPS PVLINK BUGS????

[Deanna Tran_1]

running o/s level 11i Mc/SG version 11.13
Problem : vg configure as PVlinks .
Take away the alternate path, host issue these following errors. Although iostat will show that I/O is currently using primary path to access...So it seems like the host tried so hard to access the dead path, and this path is the alternative....to the point causes the host to crash.....
After the resume of the dead path it came back to normal as this error :VM: Recovered Path (device 0x1f1f0100) to PV 0 in VG 6.
VM: Performed a switch for Lun ID = 0 (pv = 0x000000004b7a4800), from raw d


ssage repeats 21 times
VM: Recovered Path (device 0x1f1f0700) to PV 0 in VG 7.
VM: Performed a switch for Lun ID = 0 (pv = 0x000000004b7de080), from raw d

blkno: 8, sectno: 16, offset: 8192, bcount: 2048.
ssage repeats 21 times
VM: Recovered Path (device 0x1f190700) to PV 3 in VG 2.
VM: Recovered Path (device 0x1f1e0100) to PV 0 in VG 6.
VM: Performed a switch for Lun ID = 0 (pv = 0x000000004b7a4800), from raw d
VM: Recovered Path (device 0x1f180700) to PV 3 in VG 2.
VM: Recovered Path (device 0x1f1e0000) to PV 1 in VG 6.
VM: Recovered Path (device 0x1f1e0700) to PV 0 in VG 7.
Is there a new patches that i need to know for this os 11i and mc/sg 11.13 on pvlinks?
vmunix: SCSI: Read error -- dev: b 31 0x190700, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1f0000, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1e0100, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1e0000, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1e0700, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1f0100, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1f0700, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x180700, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x190700, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1f0000, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1e0100, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1e0000, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1e0700, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1f0100, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1f0700, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x180700, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x190700, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1e0100, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1f0000, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1e0000, errno: 126, resid: 2048,
vmunix: SCSI: Read error -- dev: b 31 0x1e0700, errno: 126, resid: 2048,

[Bill McNAMARA_1]

the output of
vgdisplay -v vg06
strings /etc/lvmtab
ioscan -fnkC disk
and
ll /dev/vg06/*

would help determine, although from a hunch seeing as you're on 11i, I would tend to agree with you.

I can recall working on the March 2001 11i MC OE, and seeing really strange problems with mirrordisk: It only worked if you physically pulled out the failed disk.

Try physically removing the disk.

I'd recommend installing the latest corrective bundles.
You can get them from here.. itrc.hp.com

Good luck,
Bill

[James R. Ferguson]

Hi Deanna:

Make sure that you have current LVM patches installed. One patch (PHKL_25778 (superceded by PHKL_26743)) addresses an alternate path (pvlink) problem.

"When the path of I/Os to a disk is switched from one path to another, LVM waits for the I/O requests to complete from the first path prior to completing the switch to an alternate
path. During this waiting period, resources are held (locked) in anticipation of the switch, and when the last I/O completes from the first path, some specific code is supposed to be executed to complete the switch. In the
event that a bad block is encountered on the last remaining I/O from the first path, the code to complete the switch to the alternate path is not executed. This causes LVM to hang
as it continues to wait for I/Os to finish from the first path after they have already finished".

I am *not* saying that this necessarily explains the behavior you saw, but it is worth looking at the last time applied patches.

Regards!

...JRF...

[Rudy Williams]

Deanna--

I am running much the same configuration you are (MC/SG 11.14) and see the same problems.

Have you found a resolution?

cjw