Hi,

We are using Oracle RAC 10GR2 on AIX 5.3 systems. Since RAC 10GR2 is a
clustering technique that does not use HACMP anymore for clustering
functionality, RAC itself has a deadman switch like HACMP to prevent
cluster isolation.


We had several occasions where the AIX system crashed for no obvious
reason due to the fact that the RAC deadman switch went off. The DBA
mentioned that problems with interconnect network interface, ocr and
voting device may provoke RAC to kill the AIX system.


Does anybody have experience with this, how can the Oracle timeout
setting be influenced. Oracle claims that "the settings work as
designed".
Recently, when PowerPath redundancy was used during a SAN switch
reconfiguration, all systems worked fine without any error, except the
Oracle RAC systems which ALL crashed.


Thaks in advance,


Erik

erikw_nl (AT) yahoo (DOT) com wrote:
You need to look in the log files and see why the node decided to
sacrifice itself for the cluster. Without a cause any recommendation
would just be a guess.

Daniel Morgan
www.psoug.org