Managing a cluster

Enabling negotiated failover between filers

What negotiated failover is

Negotiated failover is a feature that enables each filer in a cluster to negotiate takeover from one filer to its partner in the event of a disk shelf failure.

How negotiated failover works

When communication between a cluster is first established through the cluster interconnect adapters, the filers exchange a list of disk shelves seen on the A and B loops of each filer. If a discrepancy is found between the lists on each filer, the clustered filers perform the following negotiated failover process.

Stage	Description
1	If a filer sees that the B loop shelf count on its partner is greater than its local A loop shelf count, the local filer concludes that it is impaired because it sees fewer of its disk shelves than its partner sees.
2	The local filer asks its partner to initiate a takeover. If the partner filer is not itself impaired, the partner accepts the takeover request and asks the impaired filer to shut down gracefully.
3	The partner filer takes over after the impaired filer shuts down or after a timeout of approximately three minutes.

Enabling and disabling negotiated failover

To enable and disable negotiated failover on a cluster, complete the following step.

Step	Action
1	Enter the following command: cf nfo enable\|disable disk_shelf Use `enable` to turn negotiated failover on, and use `disable` to turn negotiated failover off. Negotiated failover is disabled by default.

Specifying time period before takeover

You can set an option to specify the number of seconds one partner in a cluster can be unresponsive before the other partner takes over.

Step	Action
1	Enter the following command: options cf.takeover.detection.seconds number_of_seconds The valid values for number_of_seconds are 5 through 180; the default is 20.

Both partners do not need to have the same value for this option. Thus, you can have one partner that takes over more quickly than the other.

Caution
If the specified time is less than 20 seconds, false takeover can occur. Use caution when assigning a takeover time of 20 seconds or less.

Determining negotiated failover status

To determine negotiated failover status on a cluster, complete the following step.

Step	Action
1	Enter the following command: cf nfo status Results The `cf nfo status` command can display the following information: Whether negotiated failover is enabled or disabled Whether a takeover is imminent due to a negotiated failover Whether a takeover has occurred due to a negotiated failover, and the reason for the takeover

Step

Action

Enter the following command:

cf nfo status

Results

The cf nfo status command can display the following information:

Whether negotiated failover is enabled or disabled
Whether a takeover is imminent due to a negotiated failover
Whether a takeover has occurred due to a negotiated failover, and the reason for the takeover