Migrating SCSI to FC path in SAN env.

[Neeraj Bajpai]

Hi Guy's

I need your expert comments on migrating SCSI device to FC device under MC service guard of two node cluster. I am explaining env. below

I have two node cluster running on LVM. These two nodes are connected to IBM storage directly via two SCSI path. Now I would like to migrate these SCSI path to FC path.

What I did, I installed one FC card on each node and shared SCSI LUN with FC HBA. So, for one device I can see three device path, two of them SCSI disk and one FC disk.

Now, my question is how can I update the cluster and LVM for new FC device.
I am new with LVM but trying to list the step here, if these step are wrong, will you guy's comment on it pl. !

1. I matched the disk serial number through STM and find out the corresponding FC device for the each SCSI devices.

2. Then remove the alternate disk from VG using vgreduce /dev/vg_name /dev/dsk/alternate_scsi_dev_name.

3. Now I will extend the VG with FC disk (eg. vgextend /dev/vg_name /dev/dsk/FC_device_name), this device will be added in a VG as an alternate disk of corrosponding SCSI disk.

4. And then I will reduce the primary path of SCSI device, so that FC device will be the primary dev. for the VG.

5. After this I need to update the cluster for any failover or resilence purpose via...
vgexport -p -s -m on one node where I did all these things...

6. Copy the map file on other node and note down the major/minor number of VG then run the vgexport for removing the /etc/lvmtab

7. Then create the group file with the same major/minor number

8. run the vgimport -s m on other node

9. Switch the package on both the node for testing purpose.


PS: All these operation will be on-line except switching the pacakage on node.
Yon can also check the cluster and VG details in attachment.


Thank & Regards

[Rita C Workman]

Morning ...

I don't have any IBM disk hardware (technology) here, but we recently switched our servers from direct attached (albeit fiber) to SAN and a simple vgexport (using the -s option as you noted) and vgimport (again the -s ) worked fine for us.

As you mention, you will need to clean up the old information afterwards. Now for myself I simple removed the old hardware path using rmsf -H to clean up /dev/dsk/c-t-d-.
See this thread & John Poff comments:
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=51818

...I do not recall doing any vgreduce....

You have already shown by your plan that you are going to redo your vgimports on the other node and test....

Just my thoughts, others I'm sure will offer more tips,
Rita

[Neeraj Bajpai]

Thanks for the link, I liked the John ans. but in your case you had two storage connected to one server for the storage migration.

But I am only changing the medium I mean from SCSI to FC media and for that my device address will be change and also I have to remove the SCSI path. That is why I have to reduce the SCSI devices from VG.


Thanks for reply.

[Sridhar Bhaskarla · Accepted solution]

Neeraj,

It's almost a day to day routine here to move the storage from one node to another though your case is slightly different. You don't have the headache of copying the data.

I understood that you are not actually migrating the data instead you are seeing the same disks through FC now.

Though your procedure works, you will still have to re-apply the serviceguard configuration if your cluster is using lock disk. Yours is a two node cluster so I assume that is the case unless you are using quorum server.

So, you have to shutdown your cluster anyway. Make it clean by deleting the cluster configuration and applying it back after importing the VGs on both the nodes.

-Sri

[Abdul Rahiman]

Hello,

Just curious what type of IBM Storage are you using ??
If you use the latest ESS storage servers(eg: shark), you would probably havve to install the IBM SDD (subsystem device driver)and it has got a whole bunch of utilities and script to do path determination, multipathing etc.. You may want to check that out. I've used it to configure my AIX and Solaris servers. I think it is sort of like the SecurePath software in the HP world.

regds,
Abdul.

[Neeraj Bajpai]

Hi Abdul, Yea I know the that using SDD, but because I have only one FC path, so I dont need to use SDD.

Sri,

Good points, honestly I skiped the quoram part in my step, two things would like to know ...

1. How do I change my Lock disk ? just editing .conf file of the cluster configuration and applying the cmapplyconf.

2. And when do I do this quoram update after applying the vgimport on both the node, then apply the cluster configuration and reboot the system.

thanks

[Sridhar Bhaskarla]

Neeraj,

1. Yes. Look for CLUSTER_LOCK_PV in your cluster configuration file and change it.
2. You can change it anytime in the ascii file but it will be effective only after applying the cluster configuration. You don't need to reboot the system.

-Sri