topic Re: ioscan false error reporting in Operating System - HP-UX

ioscan false error reporting

Richard Plourde_3 — Tue, 04 Apr 2006 14:28:41 GMT

Here is a strange one. Running ioscan on my system, I get some EMC disks that have NO_HW status. However, I look at a filesystem that is part of a vg with one of these problem disks and I can read and write succesfully to and from them.

This is an old 10.20 box.

Re: ioscan false error reporting

Steven E. Protter — Tue, 04 Apr 2006 14:47:52 GMT

Shalom Richard,

ioscan -fnk disk
ioscan -fnC disk

Same results?

I really think this is an EMC, the WWN may not be configured correctly. Could be the LUN configuration as well.

SEP

Re: ioscan false error reporting

A. Clay Stephenson — Tue, 04 Apr 2006 14:51:09 GMT

I suspect that you are actually using an alternate path.

Re: ioscan false error reporting

Richard Plourde_3 — Tue, 04 Apr 2006 15:05:37 GMT

Thanks for your reply guys,

Same result if I scan hardware or memory (ioscan -fnkC disk or ioscan -fnc disk)

I only have one HBA. So I'm not using alternate paths.

To me it looks like an ioscan bug. It is an old version of ioscan:

root@case005 [/]> what /usr/sbin/ioscan
/usr/sbin/ioscan:
$Revision: 76.1.2.1 $
PATCH/10.20:PHCO_8764 libc.a_ID@@/main/r10dav/libc_dav/libc_dav_cpe/2
/ux/libc/libs/libc/archive_pa1/libc.a_ID
Oct 2 1996 10:35:25
PATCH_10_20: ioscan.o 97/09/12
root@case005 [/]>

Re: ioscan false error reporting

Kirby A. Joss — Wed, 05 Apr 2006 11:18:54 GMT

We have seen this several times, usually as a result of a SAN event, followed by a vgscan to clean up LVM errors. I suspect that a strings /etc/lvmtab or vgdisplay -v will show that the NO_HW devices are not actually in any VG anymore. The same actual volumes are there represented by new device file names assigned when they were re-discovered and only partially recognized. On 10.20 the old device( file)s just don't go away properly with rmsf -a or rmsf -H, so you're stuck looking at old garbage.

Re: ioscan false error reporting

Richard Plourde_3 — Wed, 05 Apr 2006 12:26:22 GMT

Sorry Joss, try again.
Ioscan doesn't see the disk but the vg does. Furthermore, I can write and delete files from a filesystem belonging this this vg. The status of the lv is available/syncd.

root@case005 [/]> ioscan -fnkC disk
...
disk 17 8/4/1/0.98.27.19.1.0.4 sdisk NO_HW DEVICE EMC SYMMETRIX
/dev/dsk/c10t0d4 /dev/rdsk/c10t0d4

root@case005 [/]> vgdisplay -v vg19
...
LV Name /dev/vg19/usr2
LV Status available/syncd
LV Size (Mbytes) 12240
Current LE 3060
Allocated PE 3060
Used PV 1

...
--- Physical volumes ---
PV Name /dev/dsk/c10t0d4
PV Status available
Total PE 4315
Free PE 661

Re: ioscan false error reporting

Torsten. — Wed, 05 Apr 2006 14:50:09 GMT

It's an old system, but the patch level is old too - e.g. PHCO_8764 is dated 1996.

I would consider to patch the system or even upgrade to a 11 version.

Re: ioscan false error reporting

Sameer_Nirmal — Wed, 05 Apr 2006 16:48:05 GMT

Hi Richard,

Interesting issue against your first post at forum!

Can you tell us
1. Under which circumstances the state of those EMC disks beceame NO_HW ?
2. Which FC HBA being used?
3. Whats the versions FCMS and FC HBA device driver and their patch level?
3. As you said , the state is there for some of the EMC devices, are they Metavolumes?
4. Whats the EMC array and its firmware version?
5. Do you see any messages in syslog, dmesg
6. What you see in FC HBA stats using applcable utility e.g. fcmsutil?
7. What is the server model and 10.20 OS patch level ( core, ace )?