Can any one explain during a failover event when secondary unit takes lead as primary how it handles sessions and on going connections,is there any stateful awareness managed at any layer as switchover is a matter of few seconds as i ahve aobservered even connection/session mirroring is not enabled . or sessions replicated to HA peer too ?
This is a good question. It all really depends on the application.
By default we do not mirror any connection information. In short, afte a failover, the standby device (now active) just starts ARPing for the destination IP and it will process connection accordingly.
Most modern web application, just open a new connection to the new active device and continue as needed. So that may be why your seeing everything work pretty well during a failover.
As for timing, this depends on the failure. If it's a user induced failover, it should be pretty seamless, possibly a second or two pause. But if this is more of a network failure and the devices don't have a direct failover link, it could take up to 90 seconds. Most of the time it's about 10-30 seconds.
BIG-IP is capable of syncroniziing connection, persistence and SSL information to the peer unit so that traffic information is already present when the cluster fails over.
This requires you to setup an IP address on every device for this purpose (and optionally, a secondary mirror IP) which will be used to establish this mirroring channel.
THanks for replying . But my query is about default behaviour of F5 HA as if we havent enable mirroring at any level .How much time secondary unit takes to recover failover/will secondary unit be aware of the ongoing sessions ?
This is a good question. It all really depends on the application.
By default we do not mirror any connection information. In short, afte a failover, the standby device (now active) just starts ARPing for the destination IP and it will process connection accordingly.
Most modern web application, just open a new connection to the new active device and continue as needed. So that may be why your seeing everything work pretty well during a failover.
As for timing, this depends on the failure. If it's a user induced failover, it should be pretty seamless, possibly a second or two pause. But if this is more of a network failure and the devices don't have a direct failover link, it could take up to 90 seconds. Most of the time it's about 10-30 seconds.
