First: InterCluster LIFs only failover to the local node (hence the local-only failover policy). They should never failover to anything but local LIFs.

Second, if I recall correctly, on the early version of 8.3X (including 8.3.1) there were issues when the InterCluster LIF and/or the node-mgmt LIF was on an ifgroup. That has been fixed in 8.3.2, but it is possible the message still exists. 

I believe the release you are on is fine. Basic info:

Cluster, InterCluster & Node MGMT LIFs are all local-only (can failover, but to local node LIFs only, if available)
FC/iSCSO LIFs do not failover at all. Host-side ALUA manages failover.
All other LIFs (essentially data LIFs and vserver mangement LIFs) can and should have failover set.

On 8.3+ take advantage of the Subnets/IP_spaces/Broadcast Domains to help create your failover policies.

For instance:
Broadcast Domains: (each group are ports on the same broadcast domain that fail over to each other as needed)
Cluster (for cluster interfaces only, should include all CLUSTER ports on all nodes)
Belongs to Cluster IPSPACE
ICC (I created this for intercluster communications and include all ports used for INTERCLUSTER)
Belongs to Default IPSPACE
Management (I renamed the default to this. It includes all ports that can be used for management)
Belongs to Default IPSPACE
IP Spaces:
Cluster (for cluster interfaces only)
Contains Cluster Broadcast Domain
Default (for everything else if not multi-tenant with overlapping IP spaces)
Contains ICC & Management Broadcast Domains

Subnets:
ICC -> In my case, contains 4 IPs, one for each node. each node has an ifgroup (a0a) so, local failover is built-in
Management -> in my case, contains 8 IPs. One for each node-mgmt and vserver-mgmt



--tmac

Tim McCarthy, Principal Consultant



On Wed, Jun 29, 2016 at 8:57 AM, Joel Krajden <joelk@encs.concordia.ca> wrote:

There were warnings the last time I upgraded cDOT from 8.3.2RC2 to 8.3.2 GA and again when going from 8.3.2GA to 8.3.2P1

The warnings had no errors associate with them and the upgrade failover/giveback completed without any problems.

I wonder if the warnings are related to the fact that each node in the HA cluster has an an ifgroup using two 10G physical ports.

The ifgroup a0a is shared using one vlan for the data lifs, and another vlan for the IC lifs.

The data lifs failover to the other node's data lifs but the IC lifs failover to the local node where the target is itself.

The network infrastructure between the separate buildings where the second cluster lives is such that we cannot use another set of physical ports for the intercluster lifs on the node.

Are the vague warnings when upgrading hinting that this is not an optimal configuration?

fc1-ev::> network interface show -vserver uperc -lif email.filers -failover
         Logical         Home                  Failover        Failover
Vserver  Interface       Node:Port             Policy          Group
-------- --------------- --------------------- --------------- ---------------
uperc
         email.filers    fc1-ev-n2:a0a-539     system-defined  DATA_PUB
                         Failover Targets: fc1-ev-n2:a0a-539,
                                           fc1-ev-n1:a0a-539




fc1-ev::> network interface show -role intercluster -failover *
         Logical         Home                  Failover        Failover
Vserver  Interface       Node:Port             Policy          Group
-------- --------------- --------------------- --------------- ---------------
fc1-ev
         fc1-ev-n1_ic1   fc1-ev-n1:a0a-736     local-only      IC
                         Failover Targets: fc1-ev-n1:a0a-736
         fc1-ev-n2_ic1   fc1-ev-n2:a0a-736     local-only      IC
                         Failover Targets: fc1-ev-n2:a0a-736


--
 Joel Krajden: Core Technologies Manager
 Engineering & Computer Science Concordia University
 EV-8.142  Tel: 514 848-2424 3052
 joelk@encs.concordia.ca
_______________________________________________
Toasters mailing list
Toasters@teaparty.net
http://www.teaparty.net/mailman/listinfo/toasters