A clustered SAS Metadata Server might become unresponsive if one of its nodes cannot be contacted. The situation arises if there is no response to network communications from the unresponsive node. In this scenario, communication loss could be caused by the server being powered down or disconnected from the network or if a hardware failure has occurred.

The SAS Metadata Server reports that quorum is maintained.

This behavior is due to the SAS Metadata Server node attempting to discover the health of the missing node. The SAS Metadata Server node attempts to contact the missing node but receives no response to its request. The SAS Metadata Server waits until the network communication times out.

To date, this behavior has been reported only in an AIX operating environment.

Workaround

Reduce the AIX TCP_KEEPINIT system-wide parameter from its default value of 150 (75 seconds) to 20 (10 seconds). This change is required for all servers that run the SAS Metadata Server cluster. The parameter specifies the initial timeout value for a TCP connection. You can display the parameter's current value by using the command no -o tcp_keepinit. The value returned is displayed in half seconds.

Operating System and Release Information