Errors from SCSI driver

[Ralph Grothe]

Hi,

looks as if I've got some sort of termination issue or similar on this K450 box.

I already had a ghost mirror disk in vg02 which I suspected to have failed.
Thus I lvreduced the affected volumes from this disk and attached another to this VG.

But since then syslogd entries from the SCSI driver (like below) keep appearing.

[root@hades:/]
# grep vmunix: /var/adm/syslog/syslog.log|tail -20
Oct 19 15:29:08 hades vmunix: LVM: Recovered Path (device 0x1f029000) to PV 0 in VG 2.
Oct 19 15:29:15 hades vmunix: lbp->offset: ffffffff
Oct 19 15:29:08 hades vmunix: LVM: Restored PV 0 to VG 2.
Oct 19 15:29:15 hades vmunix: lbp->uPhysScript: 880000
Oct 19 15:29:15 hades vmunix: From most recent interrupt:
Oct 19 15:29:15 hades vmunix: ISTAT: 01, SIST0: 00, SIST1: 00, DSTAT: 84, DSPS: 00000
002
Oct 19 15:29:15 hades vmunix: lsp: 0000000000000000
Oct 19 15:29:15 hades vmunix: lbp->owner: 0000000000000000
Oct 19 15:29:15 hades vmunix: scratch_lsp: 0000000000000000
Oct 19 15:29:15 hades vmunix: Pre-DSP script dump [0000000044026368]:
Oct 19 15:29:15 hades vmunix: 0f000001 008805c1 80840000 fffffea8
Oct 19 15:29:15 hades vmunix: 7c027f00 00000000 60000040 00000000
Oct 19 15:29:15 hades vmunix: Script dump [0000000044026388]:
Oct 19 15:29:15 hades vmunix: 98080000 00000002 74072000 00000000
Oct 19 15:29:15 hades vmunix: 80840020 fffffef0 58000200 00000000
Oct 19 15:29:16 hades vmunix: SCSI: Resetting SCSI -- lbolt: 8443969, bus: 2
Oct 19 15:29:16 hades vmunix: SCSI: Reset detected -- lbolt: 8443969, bus: 2
Oct 19 15:30:44 hades vmunix: LVM: Recovered Path (device 0x1f029000) to PV 0 in VG 2.
Oct 19 15:30:44 hades vmunix: LVM: Restored PV 0 to VG 2.
Oct 19 15:33:58 hades vmunix: LVM: Failed to automatically resync PV 1f029000 error: 5

As the log entries convey, the susceptible disk is at bus 8/12

[root@hades:/]
# lssf $(vgdisplay -v vg02|awk '/PV Name/{print$NF}')
sdisk card instance 3 SCSI target 8 SCSI LUN 0 section 0 at address 8/12.8.0 /dev/dsk/c3t8d0
sdisk card instance 2 SCSI target 9 SCSI LUN 0 section 0 at address 8/8.9.0 /dev/dsk/c2t9d0


Would you agree that I should check cabling, sockets and plugs for bent pins, and termination of the attached disks?
Or do the lbolt messages suggest something else?

[A. Clay Stephenson]

Your /dev/(r)dsk/c2t9d0 is having problems but there are not enough data to pinpoint the problem. By all means check your termination, terminator power, and correct physical location of the terminators. K-box SCSI controllers have a green "TERM PWR" LED. Make sure that it is illuminated; if not, typically the on-board fuse is blown usually because of a failed or damaged component in the SCSI chain. Of course, you could simply have a failing drive. Are these bus reset recurring or was this a one-shot deal? Did someone remove a cable or other component while the box was up? SCSI bus resets are really a "normal" event and the SCSI protocol is designed to deal with them but these events should be extremely rare.

[Sanjay_6]

Hi Ralph,

Looks like the problem is with 8/8 and the disk at 8/8.9.0 (c2t9d0)

//Quote//
Oct 19 15:30:44 hades vmunix: LVM: Recovered Path (device 0x1f029000) to PV 0 in VG 2
//Endquote//

Checking the termination etc would be a good idea. Is this on a shared channel (say a SG env).

Hope this helps.

Regds

[Ralph Grothe]

I'm afraid, the resets are continous

[root@hades:/]
# grep vmunix: /var/adm/syslog/syslog.log|grep -c Resetting
169

[Ralph Grothe]

Sunjay,

you're right, 0x1f029000 obviously seems to refer to c2t9d0