RDS: ActiveBonding: Create a cluster sync point for failback
On hardware port linkups, at time the multi-cast joins fails
which delays the IP layer to bringup the interface quickly.
Subsequent multi-cast retry might succeed and then the IP
layer will be ready for IP migration. This happens very
sporadically on bare metal systems but more often on VM systems
and the number of multi-cast queries also goes up with number of VMs.
This create load of RC connection thrashing across the cluster
since the IP migration gets staggered which is not ideal for
active active. So we create a sync point so that entire cluster
gets synced up. This helps to reduce the thrashing and premature
failover attempts. Obviously its only applicable for failback
A user sysctl is provided "active_bonding_failback_ms"
in case there is a need to tune the sync point.
Orabug:
25026643
Tested-by: Michael Nowak <michael.nowak@oracle.com>
Tested-by: Dib Chatterjee <dib.chatterjee@oracle.com>
Reviewed-by: Avinash Repaka <avinash.repaka@oracle.com>
Reviewed-by: Mukesh Kacker <mukesh.kacker@oracle.com>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>