![]() |
|||
![]() |
![]() ![]() |
![]() |
![]() ![]() |
![]() |
![]() ![]() |
![]() |
| ||||||||||
Managing Switchovers and FailoversYou can trigger a switchover to swap the master and vice-master roles of the master-eligible nodes. A switchover is useful when you plan to take the master node down for maintenance. To trigger a switchover, see To Trigger a Switchover. However, if there is a problem on the master node, the master role fails over automatically to the vice-master node. In this case, the master and vice-master roles are also swapped, but because the cause is an unplanned problem, the swap is called a failover. To cause a failover, see To Reboot the Master Node Causing a Failover.
|
# /opt/SUNWcgha/sbin/nhcmmstat -c all |
The nhcmmstat command prints information on each peer node to the console window.
Log in to the master node as superuser.
Trigger a switchover.
# /opt/SUNWcgha/sbin/nhcmmstat -c so |
If there is a vice-master node qualified to become master in the cluster, this node is elected master. The old master node becomes the vice-master node. If there is no potential master, nhcmmstat does not perform a switchover.
After the switchover is complete, verify that the roles of the master and vice-master nodes have been switched.
# /opt/SUNWcgha/sbin/nhcmmstat -c vice |
If the switchover is successful, the current node is the vice master. This command also verifies that the current node is synchronized with the new master node.
Verify the cluster configuration.
# /opt/SUNWcgha/sbin/nhadm check |
For more information on nhcmmstat, see the nhcmmstat(1M) man page.
If you reboot the master node, you trigger a failover.
Log in to a peer node as superuser.
Run the nhcmmstat command to identify the master node.
# /opt/SUNWcgha/sbin/nhcmmstat -c all |
Log in to the master node as superuser.
Shut down the master node.
# init 5 |
The vice-master node becomes the master. Because one of the two master-eligible nodes in the cluster is shut down, you lose the redundancy of the cluster. To recover redundancy, restart the stopped node.
Log in to a peer node as superuser.
Verify that the vice-master node became the master node when the old master node was shut down.
# /opt/SUNWcgha/sbin/nhcmmstat -c master Executed Command: master ------------------------------ node_id = 20 [This is the current node] domain_id = 250 name = MEN-C250-N20 role = MASTER qualified = YES synchro. = NEEDED !!! frozen = NO excluded = NO eligible = YES incarn. = 1038481013 (28/11/2002 - 11:56:53) swload_id = 1 CGTP @ = 10.250.3.20 ------------------------------ |
The output shows that the vice-master node is now the master node. In addition, the new master node displays a requirement for synchronizing its disk with the disk of the old master node.
Restart the old master node, which you shut down in Step 4.
lom> poweron |
This node now automatically becomes the vice-master node.
Run the nhcmmstat command to verify that the current node is the vice-master node.
# /opt/SUNWcgha/sbin/nhcmmstat -c all Executed Command: all ------------------------------ node_id = 30 domain_id = 250 name = node30 role = IN qualified = YES synchro. = READY frozen = NO excluded = NO eligible = NO incarn. = 1038422116 (27/11/2002 - 19:35:16) swload_id = 1 CGTP @ = 10.250.3.30 ------------------------------ ------------------------------ node_id = 20 domain_id = 250 name = MEN-C250-N20 role = MASTER qualified = YES synchro. = READY frozen = NO excluded = NO eligible = YES incarn. = 1038481013 (28/11/2002 - 11:56:53) swload_id = 1 CGTP @ = 10.250.3.20 ------------------------------ ------------------------------ node_id = 10 [This is the current node] domain_id = 250 name = MEN-C250-N10 role = VICE-MASTER qualified = YES synchro. = READY frozen = NO excluded = NO eligible = YES incarn. = 1038481383 (28/11/2002 - 12:03:03) swload_id = 1 CGTP @ = 10.250.3.10 ------------------------------ |
Log in to the new vice-master node as superuser.
Verify that the node has started correctly.
# /opt/SUNWcgha/sbin/nhadm check |
For more information on the tests run by nhadm check, see the nhadm(1M) man page.
![]() ![]() |