Changed LVM physical path addresses

[Kwahae_1]

We just rebooted our servers after a power cut and having an i/o error on a lock disk in a cluster. ioscan shows the physical paths addresses have changed.

Very new to LVM. How do I change the lvmtab to reflect the changed device paths?

[Steven E. Protter]

Shalom,

/etc/lvmtab is a binary file and you don't directly change it.

Try this, but perhaps first back up vg00 with Ignite make_tape_recovery

mv /etc/lvmtab /etc/lvmtab.save
vgscan -a

That may correct the situation. More likely however the power fail has caused damaged and your lock disk is gone. Further diagnosis, dmesg /var/adm/syslog/syslog.log is probably needed.

SEP

[Kwahae_1]

Hi, I have a feeling the i/o error was due to the changed device path since cxtxdx of the lock disk is no longer part of the ioscan output.

I did a vgscan -a but did not help

[skt_skt]

Reboot does not cause the physical path changes..What is the OS version and whihc array the disk belongs to?How do you manage the multiple paths to a storage disk?

[Kwahae_1]

We have an EMC Cx7000 with Powerpath but the addresses for these did not change. The problem is on an HP VA7410 array. A range of device paths are missing from the ioscan. We are running HP-UX 11i v2.

[Mridul Shrivastava]

If you know the old device name (Old PV path)and its corresponding new device name (New PV PAth) then you could update the configuration by the following procedure.

vgchange -a n
vgcfgrestore -n -o
vgchange -a y

This will update the lvmtab file and vgconf files .... If there are multiple disk like this then you need to run vgcfgrestore for all the disks one by one.... Make sure PV paths are correct else you may get an error.

[freddy_21]

i think you must fix the va7410 problem. maybe your cluster lock at va7410.

check fc connection with fcmsutil. maybe card problem or FC cable problem.

and try to activate all volume group.
# vgchange -c n
# vgchange -a y
have any error when you activated the vg? (capture the error)


Thanks
Freddy

[whiz_kid]

take a note of ur group file information for the VGs for which you have issue.

ll /dev/*/group

mkdir /maps
cd /maps
vgexport -p -s -m .map

mkdir /dev/
mknod /dev//group c 64 0x0y0000

(replace the y with the number u hv noted from above "ll /dev/*/group" output)

then do a

vgimport -s -m .map

[Kwahae_1]

Thanks all. We just found out that the missing paths were due to faulty ports on the EVA switches.

Is it possible we could just remove the old EVA paths as they were already mirrored to the EMC CX. My problem is now how to carry out the removal. The lock is on the EVA.

[Jov]

Are the paths changed or are they missing from the VA7410?

If changed then the suggestions above will update LVM, but if they're missing (as indicated from ioscan) then you will need to resolve that and re-assign it the server first before address in the LVM side.


Jov