topic Re: HP ServiceGuard on RHEL AS 4 in Operating System - Linux

HP ServiceGuard on RHEL AS 4

AbbasNazeer — Wed, 04 Jun 2008 13:58:38 GMT

i have 2 proliant server with external MSA500 G2 SCSI Storage attached to both servers. i created a logical drive and presetned to both nodes.
when i run following commands , it works fine
to activate
vgchange --addtag $(uname -n) /dev/vgdata
vgchange -a y /dev/vgdata
TO deactivate
vgchange -a n /dev/vgdata
vgchange --delltag $(uname -n) /dev/vgdata

when created and run a package it is giving my following error and package failover is uncessful
###### Node "invigo1": Starting package at Wed Jun 4 16:36:22 AST 2008 ######
Incorrect metadata area header checksum
Attempting to addtag to vg vgdata1...
addtag was successful on vg vgdata1.
Jun 4 16:36:22 - Node "invigo1": Activating volume group vgdata1 .
Jun 4 16:36:22 - Node "invigo1": Checking filesystems:
/dev/vgdata1/lv1
e2fsck 1.35 (28-Feb-2004)
/dev/vgdata1/lv1: clean, 16/141410304 files, 4465311/282814464 blocks
Jun 4 16:36:22 - Node "invigo1": Mounting /dev/vgdata1/lv1 at /data
Jun 4 16:36:22 - Node "invigo1": Adding IP address 10.253.30.17 to subnet 10.25
3.30.0
###### Node "invigo1": Package start completed at Wed Jun 4 16:36:22 AST 2008 #
#####

####### Node "invigo1": Halting package at Wed Jun 4 16:40:25 AST 2008 #######
Jun 4 16:40:25 - Node "invigo1": Remove IP address 10.253.30.17 from subnet 10.
253.30.0
Jun 4 16:40:25 - Node "invigo1": Unmounting filesystem on /data
Jun 4 16:40:25 - Node "invigo1": Deactivating volume group vgdata1

please help me thanks

Re: HP ServiceGuard on RHEL AS 4

AbbasNazeer — Wed, 04 Jun 2008 13:59:57 GMT

also see the attached log file

Re: HP ServiceGuard on RHEL AS 4

Matti_Kurkela — Wed, 04 Jun 2008 20:26:51 GMT

Your system is producing an error message: "Incorrect metadata area header checksum". It apparently does not prevent package start, but it confuses the package halt procedure.
The actual VG deactivation is successful, but the package control script cannot confirm it.

The error message would suggest that the LVM metadata is damaged on the shared disk.

What is the output of this command?

vgck -v vgdata1

Your package seems to start completely, but then the package halts four minutes later. Did you do this manually, or did it happen by itself?

If it happened by itself, have you defined any services to the package? If there is a problem with one of the services, the package will attempt to restart the service the configured number of times. If that does not help and the service keeps crashing, then ServiceGuard will try to failover the package... but in this case, the error message about the metadata area header checksum causes SG to consider the package halt a failure.

The package halt procedure is *critical* for a successful failover. It must be completed without any errors, or the failover procedure stops, to minimize the risk of data loss.

MK

Re: HP ServiceGuard on RHEL AS 4

AbbasNazeer — Thu, 05 Jun 2008 05:06:41 GMT

Thanks for your update. i found the problem is at the time of linux booting the /dev/VGdata is activating and this /dev/VGdata is already activated on the other node.i believe this is the reason for the following message
"incorrect metadata area header checksum"
can you tell me how to deactivate VG during RHEL linux system bootup.
i did not found in the HP ServiceGuard for linux documentation.thanks

Re: HP ServiceGuard on RHEL AS 4

Matti_Kurkela — Thu, 05 Jun 2008 10:34:09 GMT

Ouch, that's kind of bad. The node definitely *must not* activate the cluster VGs if they are in use on the other node.

http://docs.hp.com/en/B9903-90054/ch05s06.html#bdchhedc

See the paragraph about the VG Activation Protection: make sure that the required /etc/lvm/lvm.conf changes are made on _all_ cluster nodes.

Whenever you install RedHat updates to cluster nodes, check if there are LVM tools updates among them. If there is, check that the lvm.conf changes are still there after the update and re-do them if necessary. These small changes are very important for the integrity of data in your cluster.

If a disk gets mounted on two nodes simultaneously and both nodes write on it, there is a risk of *massive filesystem corruption* unless you're using a cluster filesystem that's specially designed to allow that.

Been there, seen that, restored from backups.

MK

Re: HP ServiceGuard on RHEL AS 4

AbbasNazeer — Thu, 05 Jun 2008 11:19:27 GMT

i read this docs and follow the same steps but it's not working. still activating the VG at boot of linux System

Re: HP ServiceGuard on RHEL AS 4

Andrew Young_2 — Wed, 25 Jun 2008 12:03:46 GMT

Hi.

Did you have any luck resolving this issue, I am having the same problem and solving it is very annoying.

Regards

Andrew Y