Understanding giveback

Giveback process

The giveback process is described in the following table.


Stage	Description
1	An operator issues a `cf giveback` command.
2	The takeover filer updates the failed filer's disks.
3	The takeover filer destroys the virtual filer so that no more writes can occur to the failed filer's disks.
4	The takeover filer releases the disk reservations that it held on behalf of the failed filer. This returns control to the failed filer of its FC-AL A loop disks.

When the failed filer comes up, it returns to normal cluster operation, and the users of that filer experience the effects of a reboot.

Note: Giveback occurs on the takeover filer and does not depend on the state of the failed filer.

Messages indicating a successful giveback

During a typical giveback, the cluster monitor issues messages like the following on the console of the takeover filer:

filer1(takeover)> Tue Jul 21 22:37:34 GMT [filer1: rc]: Cluster 
monitor: giveback initiated by operator

Tue Jul 21 22:37:34 GMT [filer1: cf_giveback]: Cluster monitor: 
giveback started

Tue Jul 21 22:37:35 GMT [filer1: cf_giveback]: Cluster monitor: 
giveback completed

The preceding message indicates that the giveback process was successful. When the failed filer reboots, the following messages appear:

Tue Jul 21 22:38:35 GMT [filer1: cf_main]: Cluster monitor: partner 
ok

Tue Jul 21 22:38:35 GMT [filer1: cf_main]: Cluster monitor: 
takeover of partner enabled

The last message indicates that the filers are operating normally and therefore 
either filer can take over for its partner if necessary.

Note: After a giveback, the takeover filer's ability to automatically take over its partner is not reenabled until the partner reboots successfully. If the partner fails to reboot, you can enter the cf takeover command to manually initiate a takeover of the partner.

Forcing a giveback

There are two variations of the cf giveback command that force a giveback to occur:

cf giveback -f
cf forcegiveback

cf giveback -f

The cf giveback command will not perform giveback if the following long running or stateful operations are underway on the virtual filer:

parity check
disk scrub
active cifs sessions
dump or restore
kernel download
snapshot creation or deletion
quota initialization

If the cf giveback command fails because any of the above operations are running, you can force a giveback by entering the cf giveback -f command. The cf giveback -f command forces the long running and stateful operations to terminate.

cf forcegiveback

In the event that the filer panics during the giveback process, you may use the cf forcegiveback command to release the partner's identity and resources.

Caution: Using the cf forcegiveback command may result in the loss of recently written data. The command should be used only under extreme circumstances.

Part No. 210-02334
File: clust6.htm
Last updated: 04/26/99 08:11:41