IBM Books

Administration Guide


Global shutdowns and reboots of nodes with a switch

It is possible that the primary node may fence all the other nodes, including the primary backup node, if it is not the first node to shut down during a global shutdown or reboot of the entire system. Primary node takeover can also occur if the primary node is shut down and the primary backup node remains up. Running the Equiesce command before the shutdown will prevent these situations from occurring. The Equiesce command disables switch error recovery and primary node takeover. It is used to shut off the normal error actions when global activities are performed.

|If you are using the node recovery feature of the switch admin |daemon, stop the switch admin daemon before you do a global shutdown, reboot, |or rc.switch of nodes. Stop the switch admin daemon |before running the Equiesce command. On an SP Switch run the |command:

|stopsrc -s swtadmd

|On an SP Switch2 run the command:

|stopsrc -s swtadmd2

The Equiesce command causes the primary and primary backup nodes to shut down their recovery actions. Data still flows over the switch, but no faults are serviced and primary node takeover is disabled. |For the SP Switch2 you can run the Equiesce command on a |specific switch plane or on all planes. The only switch commands that are functional after the Equiesce command completes are the Eannotator, Eclock, Eprimary, Estart, and Etopology commands.

Note:
The Eclock command applies only to the SP Switch, not to the SP Switch2.

You can disable switch error recovery and primary node takeover from the command line. For example:

Equiesce

Optionally, you can use a SMIT panel to quiesce the SP Switch.

Using SMIT

TYPE
smit

SELECT
RS/6000 SP Systems Management

SELECT
RS/6000 SP Cluster Management

SELECT
Perform Switch Operations

SELECT
Switch Quiesce

See the book PSSP: Command and Technical Reference for more information on the Equiesce command.

|If you are using the node recovery feature of the SP Switch or SP |Switch2 switch admin daemon, swtadmd or swtadmd2 |respectively, restart the daemon after the global activity is complete. |To start the switch admin daemon, run the command:

|startsrc -s [swtadmd|swtadmd2]

|If you are not using the node recovery feature of the switch admin daemon, |to reestablish switch recovery and primary node takeover run the |Estart command.


[ Top of Page | Previous Page | Next Page | Table of Contents | Index ]