Auto failover does not happen in case of two successive primary and secondary cluster failures
In case of three clusters (clus1, clus2, clus3) in a GCO with steward not configured, if clus1 loses connection with clus2, it sends the inquiry to clus3 to check the state of clus2 one of the following condition persists:
-
If it is able to confirm that clus2 is down, it will mark clus2 as FAULTED.
-
If it is not able to send the inquiry to clus3, it will assume that a network disconnect might have happened and mark clus2 as UNKNOWN
In second case, automatic failover does not take place even if the ClusterFailoverPolicy is set to Auto. You need to manually failover the global service groups.
Workaround: Configure steward at a geographically distinct location from the clusters to which the above stated condition is applicable.