Lun Lock status is unknown

[kevin leong]

Dear All ,
Currently I compile the cluster script file ,and i manage bring up the cluter .

But , during i run command cmviewcl -v , i found out both server cluster lun lock status is unknown .

Please help

Thanks
Leongko

[Steven E. Protter]

Shalom,

Please post the commands you are using and the results of said commands.

SEP

[Serviceguard for Linux]

Also, please post the following info:

Distro & kernel (uname -a works)
Serviceguard version
Shared storage type
Multipath SW if any

If there is anything in the system log (/var/log/messages) related to the lock lun during cluster fomration, that portionof the log is also helpful.

[kevin leong]

Dear All ,
I compile the command is cmcheckconf -v -C /usr/local/cmcluster/conf/cmclconf.ascii -n Muatra005 -n Muatra004 and later is cmapplyconf -v -C /usr/local/cmcluster/conf/cmclconf.ascii .

And the output is successfull apply .

Current OS version is RHEL 3 with update3 .

I also checking the /var/log/message , i found out lun lock holding by node1( Muatra005) .At the same time , i also try to unplug all network cable on the node1( Muatra005) .The result is node2 (Muatra004) TOC (reboot) ,later node 1(Muatra005) following TOC .The package fail over to node 2 (Muatra004) .

Any idea ?

Thanks
LeongKO

[kevin leong]

Dear ,
The SG cluster is version 11.16 , storage MS1000 and used qlogic failover driver kit provide from HP software and driver website .

Thanks
LeongKO

[Serviceguard for Linux]

Leongko,

On the cmapplyconf, I believe you are missing info in your post here. There should be a "-L" with the lock lun designator (for example /dev/sdc1). I'm guessing that you did the command correctly but didn't list all of the information here.

Can you caputure the output of the cmviewcl -v command and post it. That will show the LockLUN information as well as some other detail.

On the logs, I was not complete in my request. Can you attach just that portion of the log that has information about the lockLun and around the time you did the test you described.

Thanks

[kevin leong]

HI Expert ,
May I know what the command to apply the cluster file .

cmapply -v -C /usr/local/cmcluster/conf/cmclconf.ascii -n Muatra005 -L /dev/sdc2 -n Muatra004 -L /dev/sda2

This command isn't right ?

Thanks
LeongKO

[Serviceguard for Linux]

You have:
cmapply -v -C /usr/local/cmcluster/conf/cmclconf.ascii -n Muatra005 -L /dev/sdc2 -n Muatra004 -L /dev/sda2

it should be:
cmapplyconf -v -C /usr/local/cmcluster/conf/cmclconf.ascii -n Muatra005 -L /dev/sdc2 -n Muatra004 -L /dev/sda2

Again, this is likely a typo.

Also, it is unusual to have different disk identifiers for the two different nodes. Please check very carefully that /dev/sdc2 on Muatra005 and /dev/sda2 on Muatra004 are the same LUN.

Are you giving the exact command you used or are you typing an example? /dev/sda is almost ALWAYS and internal disk and could not be used as LockLUN.

Again, if you cut & paste the result of cmviewcl -v, as well as the log information requested, it would help.

[kevin leong]

Hi ,
The /dev/sda2 and /dev/sdc2 is a same lun .The device path is belong share disk so that configure as Lun Lock .

Posibble that i used the devlabel -u (device_id) -s /dev/lun on the both server .
Later on the cluster script Lun Lock change to /dev/lun ,isn't work ?

Thanks for advice
LeongKO

[Serviceguard for Linux]

The only supported device names for Lock Lun are those that are naturally defined by the drivers. So for a FibreChannel device a LockLUN must have a name in the form /dev/sdxxx. This would be teh case for your array. CCISS names are also supported with the MSA500.

Hope this helps your understanding.

At this stage, it might be faster if you put in a call to HP support. They can help get all of the information, such as the log information, necessary fromt eh system to help with your problems.

[kevin leong]

Hi ,
If I adding/removing the HDD to this system ,
the hdd device path will changing .Is mean /dev/sda2 become sdd2 .

How could I prevent on this case ? On redhat only the devlabel to fix the device path ,and SuSE enable the boot.scsidev .Any alternative solution ?

Thanks
LeongKO

[Stephen Doud]

Disable Device Mapping software on any disk governed by Serviceguard.

[kevin leong]

HI Stephen ,
May I know that how to disable the device mapping ?

Thanks
LeongKO

[Oliver Schwank]

Hello,

we are having the same problems.

We are planning a cluster with 40 LUNs and it could be possible that one day LUN 24 will be removed and all device names above will change.

It would be helpfull if we could use the device names from the linux persistent name project unter /dev/disk/by-id is there any way to do that?

Bye

Oli

[Stephen Doud]

Kevin,
I'm afraid I can only offer this as anecdotal experience, having worked with a customer that had configured the lock lun device to use device mapping.

I think the 'udev' command may be helpful, but I'm not certain.

[Serviceguard for Linux]

udev should help. Be aware that the LockLUN name must be of the form /dev/sdxx for FC devices or /dev/cciss/. The lockLun code checks the actual devices and uses the name to determine how to do the test.

The format of the name should not matter for Luns that are used with LVM and packages.