LVM vg[] POWERFAILED

[j773303]

I see the message in dmesg, does anyone know what's happend?

LVM: vg[5]: pvnum=0 (dev_t=0x1f041100) is POWERFAILED
LVM: Recovered Path (device 0x1f041100) to PV 0 in VG 5.
LVM: Restored PV 0 to VG 5.

[Jarle Bjorgeengen]

Could be a poor connection to one of you disks.

Most likely hardware-related, anyway.

Rgds Jarle

[Bernhard Mueller]

Hello,

basically LVM thought one of your disks/luns was no longer there, then it came back and recovered the old status.

If this happens frequently, check connections to the disks / disk array, if you are using EMC or the like you may need the change the timeout value for the Luns by
pvchange -t 180

Regards,
Bernhard

[Massimo Bianchi]

Hi,
this means the the first disk in your fifth 5 had a malfunction, but it was momentary.

How to know the fifth vg: do a "strings /etc/lvmtab | grep vg" , the 5th in the list is yours.

Have also a look at your "dmesg" output.

HTH,
Massimo

[Ian Dennison_1]

You have a possible pending disk failure. If it is part of a Storage Array (XP, EMC) call the vendor.

If it is part of a SAN, investigate the status of the connection.

Below are some useful forum entries for troubleshooting problems like this.

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x80827d4cf554d611abdb0090277a778c,00.html

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0xc291e7613948d5118fef0090279cd0f9,00.html


http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x667e7cffe265d711abdc0090277a778c,00.html

Share and Enjoy! Ian

[Keely Jackson]

Hi

It means that the system temporarily lost sight of the disks. If these are in dmesg then they may not necessarily be recent. Also worth checking syslog or EMS if it is running.

If they are connected via fibre-channel it is worth checking your patch state.

It is also worth increasing the timeout parameter for the physical volume. The default is 30s (I think) but it is ok to up it to 180 and see if the problem goes away: eg

pvchange -t 60 /dev/dsk/c4t1d1

If the messages persist, you may have a problem with the disk.

Cheers
Keely

[Cheryl Griffin · Accepted solution]

I was advised by one of the senior hardware engineers recently that you should not go straight to a pv timeout of 180 or higher. LVM does default to 30, so you want to bump up the timeout gradually. Try 60-90 first. You want to go with the smallest increment that may resolve these messages (if these messages are not associated with an actual HW failure).

The logic behind using the smallest increment is that the timeout affects the reporting of the disks. If there is a problem, you want the quickest response back.

In addition, you want to check for the latest SCSI & LVM patches. For instance SCSI patch PHKL_18543 for 11.00 (and it's dependencies).

The problem disk should be c4t1d1:
# ll /dev/* |grep 041100

[Dario_1]

Hi!

Also try the following to see if you get any errors:

dd if=/dev/rdsk/cXtYdZ of=/dev/null bs=1024

Regards,

Dario

[Dario_1]

Hi!

Also try the following to see if you get any errors:

dd if=/dev/rdsk/cXtYdZ of=/dev/null bs=1024

where cXtYdZ should be c4t1d1
Regards,

Dario

[Steve Post]

I have an open hw case on this problem right now.....

dev_t=-1f041100 is..... c4t1d1.

vg[5] means is is the volume group with
/dev/vg???/group with a nodeid of 5.

If this is hooked to an EMC array, they (EMC) will probably tell you to increase the timeout to 180.
i.e. pvchange -t 180 /dev/dsk/c4t1d0.
What is the timeout value set to now???
run pvdisplay -v /dev/dsk/c4t1d0 | more

If you run "stm" to see the error logs, you may be surprised to find this error is occurring more than you think.

Also vi file /var/adm/syslog/syslog.log. In there you can see the time of the error.

Is this happening at the same time every day? Perhaps during a huge backup?

I can't tell you what's causing it. I wish I knew. That's why I have a call into hp on it. (and EMC).

Steve

[Steve Post]

one more big thing. You should know that phkl_18 is called "the patch from hell", and the "line-in-the-sand" patch.

You can't uninstall it.
You can't reinstall it.
If you have the patch on your system, leave it alone.
Search on the patch in the forums, you'll see what I mean.