Here I’ll describe what will happen if your sysadmin are going to restart VIO servers and not co-ordinate activities with DBA.
Although you have multipath configured, VIO servers are critical for normal IBM AIX LPAR functionality, as they share HW devices to LPAR VMs.
If you are having RAC or RAC One Node configuration, this is what you’ll find in GI log file:
[cssd(7078064)]CRS-1612:Network communication with node rtl-mom-db-1p (1) missing for 50% of timeout interval. Removal of this node from cluster in 14.905 seconds
[cssd(7078064)]CRS-1611:Network communication with node rtl-mom-db-1p (1) missing for 75% of timeout interval. Removal of this node from cluster in 6.863 seconds
[cssd(7078064)]CRS-1610:Network communication with node rtl-mom-db-1p (1) missing for 90% of timeout interval. Removal of this node from cluster in 2.841 seconds
[cssd(7078064)]CRS-1609:This node is unable to communicate with other nodes in the cluster and is going down to preserve cluster integrity; details at (:CSSNM00008:)
in …/cssd/ocssd.log.
In case of RAC, you will end with node eviction of LPAR where you restarted VIO server.
Restarting VIO server on RAC One Node configuration will result with production downtime and reallocation of services on the passive node.
If you have single instance database (either Oracle Restart which is ASM based or simple single instance based on file system), users connections will reset, and will need to reconnect.
The following wait event will show up in the AWR report:
SQL*Net break/reset to client
and this is what you’ll find in AIX error log:
LABEL: SDDPCM_PATH_RECOVER
IDENTIFIER: 67150733
…
Description
PATH HAS RECOVERED
…
Probable Causes
ADAPTER HARDWARE OR CABLE
DASD DEVICE
…
Failure Causes
EF00
In this case, although there are more than one VIO server, this is not going to prevent issues that will be caused by this particular action.
If you are aware of possible consequences, you will plan VIO server restart for some quite period when the impact will be minimal.
For that reason maintaining a open communication between sysadmins/storage admins/DBA/network admins and business, as well as involving people with different skills is crucial to perform any kind of action in complex enterprise environment.
Comments