read i/o error

[Sanjay Dholakiya]

I am a new hp-ux sysadmin, I am facing a problem and need some help.

We have an import running, which was stuck(hung), in dmesg output read i/o error was reported due to POWERFAILURE on SCSI device.
I checked syslog, and same error message is continueing.

Error message :
Jul 17 06:36:48 eagle vmunix:SCSI: read error -- dev b: 31 0x042000, errno:126, resid:2048
blkno:8, sectno:16, offset:8192, bcount: 2048,
LVM:vg[2]:pvnum=0 (dev_t=0x1f042000) is POWERFAILED
LVM:Recovered Path (device 0x1f042000) to PV 0 in VG 2.
LVM:Restored PV 0 to VG 2.

I have still not checked device is plugged correctly or not but this is new refurbished N-CLASS HP server we've received and I have started configuring.

Could somebody shed some light of experience on this problem ?

Thanks in advance for your help.

[Pete Randall]

Sounds like you're having problems with a disk in vg02. Check ioscan, vgdisplay, pvdisplay output, as well as /var/adm/syslog. Possible a failed drive.


Pete

[Stefan Farrelly · Accepted solution]

Even though your server is refurbished it doesnt mean it has new disk drives. Theyre used, and one is failing. Its either powering down (sure sign its about to die) or developed a bad block needed a rebuild. I would get HP to replace it. The device in question is listed in the error;
ll /dev/dsk | grep 042000

[PIYUSH D. PATEL]

Hi Sanjay,

What happened here is that one of the disk in the Volume group 2 lost connectivity to the power supply and after sometime the connectivity was restored. It could be due to loose connection or bad disk. The disk is also giving you I/O errors and there may be bad blocks in the disk. If these errors are comming frequently then you need to change the disk.

HTH,
Piyush

[Dario_1]

Hi!

If you have a service contract, call HP and have them replace the disk.

Regards,

Dario

[Pete Randall]

If you don't have a service contract yet, you need to talk to the supplier that furnished you this equipment. They *should* set it right for you.


Pete

[Elena Leontieva]

You may want to do:

vgdisplay -v vg02 ( to see if all disks are "available");

diskinfo /dev/rdsk/c4t2d0 - if it returns valid information;

If any of the above fails - replace the disk.
You should run some additional commands to replace LVM drive.


Elena.

[Enrico P.]

Hi,
if you have an high disk activity should be that the disk in the question was in time-out.
You can try to change the IO_timeout parameter with the pvchange command.
See man pvchange option -t.
A good value is 120 sec.
I know the default is 60 sec.

Enrico.

[Tim D Fulford]

Hi

I just wanted to chim in & show you how you could come to the above conclusions yourself

Jul 17 06:36:48 eagle vmunix:SCSI: read error -- dev b: 31 0x042000

This means device with major number 31 & mimor number 0x042000. is failing (SCSI error).

You can track the major number to a device/driver by doing lsdev | egrep 31. I think this is a disk devcice, dsk or rdsk. so cd /dev/dsk (or maybe /dev/rdsk) & do ls -l, track down device with minor number 0x042000. This is the device that is failing.

Also the LVM:... messages are also a give away

Regards

Tim

[Sanjay Dholakiya]

Thanks Pete, Stefan, Piyush, Dario, Elena,Enrico and Tim for your suggestions. Now, I will do some homework and take it to HP.

[Eugeny Brychkov]

Reported device is /dev/[r]dsk/c4t2d0.
I think it's not internal drive (as soon as bus is 4 and VG is 2).
Please check if there're more disks/devices on this SCSI bus (beginning with c4tXdY). Check if everything is connected securely. Check if there're such events for other connected disks. Check disk's readability with dd:
dd if=/dev/rdsk/c4t2d0 of=/dev/null bs=1024k
if finally it will return 'I/O error' then disk could be defective
Eugeny