HOW WE CAN TRANSFER A FILE SYSTEM FROM ONE NODE TO ANOTHER NODE WHILE USING SERVICE GUARD CLUSTER?

[SANOJKUMAR.C.V]

HOW WE CAN TRANSFER A FILE SYSTEM FROM ONE NODE TO ANOTHER NODE WHILE USING SERVICE GUARD CLUSTER?SUPPOSE ONE NODE IS DOWN.

[Sajjad Sahir]

if one node is down no need to trnasfer a file system u have to up the node which is down.

thanks and regards

Sajjad Sahir

[Syed Muhammad Ali Haide]

Dear Sanoj,

clarify your question? the other node was the part of the cluster?

Regards,
Ali.

[DeafFrog]

Sanoj ,

Refer to man pages of vgexport / vgimport .

[Matti_Kurkela]

> HOW WE CAN TRANSFER A FILE SYSTEM FROM ONE NODE TO ANOTHER NODE WHILE USING SERVICE GUARD CLUSTER?SUPPOSE ONE NODE IS DOWN.

Technically, Serviceguard does not really "transfer" filesystems - it controls access to shared storage by restricting VG activation.

To "transfer" a filesystem, that filesystem must already be:
- physically located on shared storage that is presented and accessible on both nodes
- located in a VG that is in clustered mode (vgchange -c y) if LVM is used; VxVM uses different procedures
- the VG and filesystem should be included in a Serviceguard "package".

If the package was running on a node that went down, Serviceguard should have automatically moved that package (including the VG(s), filesystem(s) and IP address(es) belonging to that package) to the other node IF the package's AUTO_RUN attribute was enabled (cmmodpkg -e ) and the other node was eligible to run the package. The eligibility depends on configured package attributes.

If AUTO_RUN was disabled but everything else is OK, the package is capable of being transferred but does not transfer automatically.

To transfer it manually, first run "cmviewcl" to verify the current state of the package. It should be down, but the cluster should still be up. Then, run "cmrunpkg " on the node that is up. If the Serviceguard package is properly configured, this command should automatically activate the necessary VG(s), mount the filesystem(s), activate the IP address(es) and start any necessary application processes.

If this command is not successful, read the error messages, package log files and the system log to find out what went wrong.

If "cmviewcl" reports the entire cluster is down, you'll have to start it first. Normally you need _more than half_ of the nodes to start the cluster; if you have a two-node cluster with one node down, you must use "cmruncl -n ".
If you do this, it will be YOUR RESPONSIBILITY to make sure the other node is really down and that nobody will run a similar command to start another copy of the same cluster on the other node when it is fixed (the other node must join to the already-running cluster using "cmrunnode", not set up a second copy of the cluster with another "cmruncl").

If the filesystem is not configured into a Serviceguard package, then Serviceguard cannot really help in transferring the filesystem: you must use the same procedures you would use without Serviceguard.

If the filesystem is located on disks that are not shared, you have to either fix the node that is down, or make the disks accessible to the node that is up (i.e. physically move disks, connect cables etc.). Serviceguard cannot magically transfer data out of hardware that is down.

MK

[Stephen Doud]

If the file system mount is controlled by a Serviceguard package, and that package is enabled to run on the remaining node, you should be able to start the package and have Serviceguard mount the file system on the active node.
If you don't know whether the file system is controlled by a Serviceguard package, use 'cmviewcl -v -f line | grep