2010-07-29 03:56 AM
I have setup our fsa2020 in active active configuration.
Both onboard controllers are setup in the same way lacp vif to bond the two network ports. I have tested removing one cable from a controller and the network connection stays live as expected.
If I pull out both network cables on controller A the connection dies at this point with clustering enabled I though controller B would take over. From the Filer view, manage cluster controller B states that controller A is up. If I force a failover to controller B everything work all shares etc are available.
Should the cluster not see that the network connection to controller A is down and take over automatically? If so how do I configure this?
2010-07-29 11:55 AM
This is not the case by default, but it can be set if desired. The original idea was to use failover virtual interfaces to handle this case instead of a failover, however if you want to force a failover in the case of an interface failover (physical or virtual), then set the following option to on:
This option allows negotiated takeover to be enabled when the HA nodes detect failures in network interfaces. Only those network interfaces that have explicitly enabled negotiated failover via the ifconfig command will be monitored. By default, this option is set to off.
This option is available only when cf is licensed and changing the value on one filer automatically changes the value on the partner filer.
cf.takeover.on_network_interface_failure.policy This option determines what policy to apply for triggering negotiated failover when network interfaces fail. There are two policies that are currently supported: all_nics implying failover when all network interfaces participating in negotiated failover fail and any_nic implying failover when any one of the network interfaces participating in negotiated failover fails. By default, this option is set to all_nics.
Then as stated in the first option you will need to set the appropriate interfaces (do the vif if you are using vifs) to negotiated failover using ifconfig. This will need to be reflected in /etc/rc as well.
Hope this helps.