Alletra 6030 controller failover conditions

[DelMagna]

Hi

I have a question if the following scenario will result in a controller failver on an Alletra 6030 running the latest firmware versions.

The scenario is for an Alletra 6030 array that as controller A as the active controller. The array is configured for iSCSI storage presentations to ESXi hosts. There are two iSCSI VLANs and each controller has an iSCSI port with a single DATA ip address on each VLAN. Each iSCSI VLAN also has a discovery IP address. Therefore each controller has just a single iSCSI port configured for each iSCSI VLAN - this is eth1a for iSCSI VLAN-1 and eth3a for iSCSI VLAN-2. 

If either eth1a or eth3a suffers a failure (perhaps SFP failure) on the ACTIVE controller, will the array initiate a failover to the passive controller or not? The passive controller has not suffered the same interface issue.

many thanks

Del

[buzzsubash]

Hello,

That is correct, it will trigger a failover.

The logic in Nimble OS is designed in a way that, the controller with more number of active connections will try to be "active" all the times.

[DelMagna]

Hello

Many thanks for the info and I think I read that somewhere - however, does that only apply to DATA interfaces? It's not behaviour I have seen with the management network. We have both Eth0a and Eth0b configured on both controllers and when one of these is removed on the Active controller a failover is NOT triggered. For management it seems that only when both management ports fail will the controller then failover.

That also leads me to a second question. If we need to manually perform some 'planned maintenance' that will result in a single disconnected iSCSI interface on the Active controller (eg re-route an Alletra iSCSI cable), to prevent a controller failover, should we first manually failover the controller rather than trigger a automatic failover? Is there any advantage of doing this?

OR should we first remove the equivalent iSCSI cable from the Passive controller to leave this passive controller with fewer 'online' interfaces than the Active controller. Thus when we then pull the cable on the Active controller (to re-route it), a controller failover will NOT be triggered as the passive controller would not have more active ports than the Active controller.

thanks in advance

[buzzsubash · Accepted solution]

It should be applicable for interfaces in Managment network as well, and an interface will be reported as down only after the cable disconnected for more than 5 seconds. Please note, eth0b or the secondary management IP to be assigned only if there is a second array in same group. But let me see if I can test this scenario in our lab array.

And to answer the question, both steps should work fine and do not have any added advantage. 

[DelMagna]

Many thanks. I’ll look forward to hearing about the findings from your lab tests. Many thanks