How much latency is there for planned and unplanned failover?

[Former Member]

Planned failover is the one in which the administrator of SAP HANA shut down the system for some reason pertaining to maintenance, patching data services, etc.

Unplanned failover is very abrupt in nature and is caused by software or hardware failure, natural or man-made disasters.

What i feel is that there are following steps involved in the failover and these make up for the total latency:

1) Identifying that the system failure has occurred

2) trying to bring back the system to its normal stage with minimum efforts

Under the case of planned failover we can notify the system about the reason of the downtime, which in turn eliminates system's effort to recognize the problem. So we are eliminating this latency. Now the system have just to take the appropriate step to complete failover process.

In case of unplanned failover the system itself to recognize the reason for the downtime and this adds up to identifying the cause latency. After identifying the cause the system takes in the appropriate step to complete failover process.

I believe what i am assuming is correct. And if i am correct can anyone help me evaluate the actual latency for planned and unplanned failover? Feel free to add onto this discussion

[fabian_herschel]

For SAP HANA in system replication you might like to view my SCN article:

http://scn.sap.com/docs/DOC-56278

And my Blog: http://scn.sap.com/community/hana-in-memory/blog/2014/04/04/fail-safe-operation-of-sap-hana-suse-ext...

These articles do not answer the question about times, but they explain the concept of a already available solution to optimize down time in take over situations or to configure local restart.

In both cases the Pacemaker resource agents already available in SUSE Linux Enterprise Server for SAP Applications 11 SP3 could automize the take over.

For "normal" SAP HANA internal HA you do not need the resource agents described in my article. Its designed for system replication.

The benefits comes mostly from the table-preload. This optimizes the database load time dramatically. So I really like SAP for implementing this HANA feature, while we (SUSE) automaized the takeover in addition with the correct placement of the "virtual" IP address for the access of DB clients.