What is the expected response and recovery time for VIP's in CSS 11503

vijay.gaganam · ‎04-22-2008

What is the expected response and recovery time for VIPs and Gateways configured in the load balancers to be available?

During switch over, connected equipment was unable to contact gateways configured in load bala ncer for extended time. CSS11503 detected uplink loss causing a failover to standby load balancer. The gateways built in the load balancer were unresponsive after the failover for approximately 15 minutes. There was a period of approximately 15 minutes where one of the gateways in the load balancer was unavailable. This caused our NFS clients to time out and our software to force a restart due to unresponsiveness.

Gilles Dufour · ‎04-23-2008

it all depends on spanning-tree but normally it should never take more than 1 minute.

If you say 15 minutes, it would be a problem of arp entry that was not updated with a GARP and only after a arp timeout (which could very well be 15min) did the gateway recognize the CSS failover.

You'll need to test a failover and capture sniffer trace on the link between CSS-Gateway and see if a GARP is sent by standby CSS when it takes over.

Gilles.

Report Inappropriate Content · ‎11-27-2008

Hello,

I have the same problem with two CSS in box-to-box redundancy. In my case, the gateway is a FWSM, after the switchover the FWSM don't update the ARP table with the new mac-address of the standby CSS. It seems that the FWSM don't accept the GARP.

The software version for the FWSM is 3.1(7).