There are four tunable parameters in this algorithm.
inactive_time
ping_timeout
repeat_test
slow_network
These parameters provide an adjustable trade-off between speed and correctness of fault detection. See Table 5-3 for more information.
Use this procedure to change the default Public Network Management (PNM) values for the pnmd(1M) daemon.
Become superuser on any node in the cluster.
If it doesn't already exist, create the pnmparams file.
# vi /etc/cluster/pnmparams |
Use the following table to set PNM parameters.
Settings in the /etc/cluster/pnmparams file apply to all NAFO groups on the node. Lines started with a pound sign (#) are ignored. Other lines in the file must be of the format: variable=value.
Parameter |
Description |
---|---|
inactive_time |
Number of seconds between successive probes of the packet counters of the current active adapter. Default is 5. |
ping_timeout |
Time-out value in seconds for the ALL_HOST_MULTICAST and subnet broadcast pings. Default is 4. |
repeat_test |
Number of times to do the ping sequence before declaring that the active adapter is faulty and failover is triggered. Default is 3. |
slow_network |
Number of seconds waited after each ping sequence before checking packet counters for any change. Default is 2. |
warmup_time |
Number of seconds waited after failover to a backupadapter before resuming fault monitoring. This allows extra time for any slow driver or port initialization. Default is 0. |
The changes don't take affect until the next time the pnmd daemon starts up.
The following shows a sample/etc/cluster/pnmparams file, with two parameters changed from their default values.
inactive_time=3 repeat_test=5 |