Cluster Failover

Unanswered Question
Sep 16th, 2009

Hi,

is there any possibility to increase HA speed of the two Fabric Interconnects?

In our redunancy tests it took about 100 seconds to switch the subordinate to the primary - during this time the cluster management ip adress won't work.

is there a hidden/undocumented command to decrease the time between the checks?

Or how do you perform these tasks?

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
David McFarland Wed, 09/16/2009 - 07:19

If this is with respect to UCS, switchover between the primary and secondary UCS 6120 will result in a control plane outage. The time for the control plane to recover is a funtion of the size of the managed environment. It will take longer with more managed devices.

Control plane failover affects the ability to perform active management of the UCS. It does not affect the data plane.

Unless, there is an actual switch reload or total failure. Then the data plane switch over can take several seconds. Impact to hosts will depend on whether A-B, B-A HA is configured, or the NIC teaming, or, if SAN dual fabric with multipathing is configured.

In short, there are no commands available to decrease the UCS managment recovery time.

Actions

This Discussion