topic Re: SCSI: write error - explained in Operating System - HP-UX

SCSI: write error - explained

D Block 2 — Sun, 02 Oct 2005 22:54:48 GMT

what does SCSI: error in syslog imply ?

Do you get these errors ? And how to correct ?

details and concerns
----------------------------------------------

what does Error message tell me ?
errno: 126 ?
resid: 4096 ?

grep SCSI: /var/adm/syslog/syslog.log:

Sep 22 09:39:51 MYSERVERNAME vmunix: SCSI: Async write error -- dev: b 31 0x043300, errno: 126, resid: 4096,

kernel's fprintf() format might be this:
# pwd
/stand
# grep SCSI: vmunix | grep -i 'write'
SCSI: Async write error -- dev: %s %d 0x%06x, errno: %d, resid: %d,

I can find the 'dev:' from ls -l /dev/dsk.. and match to the actual special disk device.

Re: SCSI: write error - explained

RAC_1 — Sun, 02 Oct 2005 23:02:59 GMT

The device that is giving you problem is 0x043300

b 31 - it is disk device. You may have disk problem.

ll /dev/dsk/* | grep 0x043300
This shold be disk c4t3d3
Check that disk - pvdisplay, diskinfo and dd test
dd if=/dev/rdsk/c4t3d3 of=/dev/null bs=1k

Re: SCSI: write error - explained

Mugilvannan — Sun, 02 Oct 2005 23:06:04 GMT

http://forums2.itrc.hp.com/service/ forums/questionanswer.do?threadId=634395

Re: SCSI: write error - explained

Adisuria Wangsadinata_1 — Sun, 02 Oct 2005 23:09:33 GMT

Hi,

The device that have a problem is c4t3d3, check with 'ioscan -nfC disk' whether the status become NO_HW or missing or else.

Run the disk exercise with dd command :

# dd if=dev/rdsk/c4t3d3 of=/dev/null bs=1024

Check also with :

# pvdisplay -v /dev/dsk/c4t3d3
# diskinfo -v /dev/rdsk/c4t3d3

Hope this information can help you.

Cheers,
AW

Re: SCSI: write error - explained

Arunvijai_4 — Sun, 02 Oct 2005 23:12:58 GMT

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=724357
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=849572

You might have a corrupted disk, try RAC's method # dd if=/dev/rdsk/c4t3d3 of=/dev/null bs=1k

-Arun

Re: SCSI: write error - explained

Stephen Keane — Mon, 03 Oct 2005 04:41:14 GMT

The 'resid' is the residual bytes left after the data transfer was attempted.

For example if the I/O routines were trying to write, say, 8k of data to the disk and only half was transferred then the resid would be 8192 (i.e. 8k) - 4096 (number of bytes actually transferred) = 4096.

Or, say the I/O routines were trying to transfer 4k to the disk, and none of it was successfully transferred then the resid would be 4096 (attempted) - 0 (successfully transferred) = 4096 (resid).

You would normally expect the resid to be zero. The errno is an indication of why the data transfer could not be completed. It would be 7E in hex and is a vendor specific SCSI error indicator.

Re: SCSI: write error - explained

D Block 2 — Mon, 03 Oct 2005 15:07:14 GMT

Thanks guys for the quick response.

Mugilvannan, can you send that link again.

Re: SCSI: write error - explained

Nguyen Anh Tien — Mon, 03 Oct 2005 20:48:05 GMT

The error no 126 is a :
errno = 126 (EPOWERF) means the SCSI driver could not complete the read/write requests within specified timeouts.
The errors listed above are occurring in the syslog indicating timeouts have occurred but there are no vg[#]: pvnum=# (dev_t=##) is powerfailed as normally seen.

Re: SCSI: write error - explained

Andrew Merritt_2 — Tue, 04 Oct 2005 02:35:50 GMT

Also, if you've got the EMS HW Monitors (part of the OnlineDiag package) installed, you should have some events logged in /var/opt/resmon/log/event.log which may give more detail about the problem.

Andrew

Re: SCSI: write error - explained

D Block 2 — Fri, 07 Oct 2005 21:13:16 GMT

have you used the FCP Ping ?

http://www.santools.com/smart/SMARTMONUX-Manual.pdf

1.19 Storage Area Network (SAN) Device Ping
This function can be equated with a standard TCP/IP ping. It is used to both determine connectivity to a device, and
to report the amount of milliseconds it takes for a packet of data to get to the device and be acknowledged by it.
Syntax
smartmon-ux -fcping PortWWN LUN_Number [Attempts]
The LUN_Number would typically be zero for standard disks and tapes. It is quite common to be non-zero for logical
disks created by external RAID subsystems.
The PortWWN corresponds to the fibre channel port WWN for the selected device. This information can be obtained
by a variety of methods, including:
Â· Running smartmon-ux -fc which will dump all port and WWN info for the devices it can see
Â· Your HBA management software
Â· Your operating system's registry or boot logs (i.e, /vary/log/messages or dmesg)
The optional attempts field is used to tell the program how many attempts it should make. If you enter zero, then it
will send data indefinitely, or until you abort or kill the program.
Example
D:\TEST>smartmon-ux -fcping 22:00:00:20:37:E6:0A:38 0
SMARTMon-ux [Release 1.23, Build 07-DEC-2003] - Copyright 2003 SANtools, Inc.
http://www.SANtools.com
Port 22:00:00:20:37:E6:0A:38 replies in 0.010s as SEAGATE ST336605FC
Port 22:00:00:20:37:E6:0A:38 replies in 0.000s as SEAGATE ST336605FC
Port 22:00:00:20:37:E6:0A:38 replies in 0.000s as SEAGATE ST336605FC
59

Re: SCSI: write error - explained

D Block 2 — Fri, 07 Oct 2005 21:16:05 GMT

If the disk is responsing slow, try running the "sar -d" command to see the 'avwait' times.

I have also learned that the Fiber cable could be an issue.. we change-out the cable to the swith.. to the server.. this might help redude the "avwait" time and also reduce the SCSI: write errors..