Est Avg Time for MC Service Guard to Failover a Failed Node to 2nd Node and/or Phy Standby

[yakub.dba.architect]

Please share your experience with the following:

Given a 2 node (node-1 is Active and node-2 is passive) MC Service Guard RAC with Oracle What is the average/estimated time for MC Service Guard to detect a Failed Node and then fail over to the Passive Node. How much time could this take in minutes? Thanks

[Steven E. Protter]

Shalom,

It depends is the only valid answer you can get here. For real data, you will need to run a test.

Let me run through what MC/SG does when a fault is detected.

It activates a volume group on shared storage.
It runs package scripts, which start applications up. This is the time consuming step. If you are starting an Oracle database with access to 500 GB of data, its going to take a little wile to get started.

SEP

[Viktor Balogh]

SEP has right. But add the time of failure detection too. This depends on the settings of the following Serviceguard parameters:

if you are using a quorum server:

QS_POLLING_INTERVAL
QS_TIMEOUT_EXTENSION

HEARTBEAT_INTERVAL
NODE_TIMEOUT
AUTO_START_TIMEOUT
NETWORK_POLLING_INTERVAL

All the values are in microseconds, for further info look for your cluster ascii file and read the comments.

[Viktor Balogh]

...oh, and by the way: you can be only sure after testing.

[yakub.dba.architect]

Thanks,

With using HP's MC ServiceGuard, I understand that I will not be using Oracle Clusterware because MC SG manages the cluster. What other Oracle components will I need other than Oracle RAC database.

[Viktor Balogh]

you need Serviceguard Extension for RAC, here you are:

http://h20338.www2.hp.com/enterprise/w1/en/os/hpux11i-serviceguard-sgerac.html