Ping nodes with MPIO - what are people using?

[albrow]

Thanks to Todd and Janusz, we've managed to set up our two DSS V6 boxes using iSCSI Failover with MPIO, talking to 3 VMware ESX4 hosts.

Our DSS V6 nodes are configured like this:

NIC1 - iSCSI network 1 (VLAN 201)
NIC2 - [BOND0] Replication (no VLAN - direct connection)
NIC3 - iSCSI network 2 (VLAN 202)
NIC4 - [BOND1] Management (VLAN 101)
NIC5 - [BOND0] Replication (no VLAN - direct connection)
NIC6 - [BOND1] Management (VLAN 101)

NIC1 and NIC3 have both a physical and virtual IP (for failover).

Open-E support have told me that DSS V6 only supports one ping node. (I've actually tried using more than one ping node by entering 2 IPs separated by a colon. All that happened was a dramatic slowdown in the speed of the primary node and a nice red "fail" label in iSCSI Failover.)

However, because we're using MPIO, our iSCSI traffic travels through 2 NICs and therefore 2 subnets. As a result, we surely need to have a ping node on both subnets. Otherwise, if we unplug one of the network ports on the primary, or if the switch dies, then there is a risk of a failover - something that's going to be more hassle than it's worth.

One way of getting around the problem would be to put the ping node on one of the subnets with bonded NICs. However, this isn't representative of whether the iSCSI subnets are up or not. For example, the iSCSI nodes will function fine if the management network was to go down.

What do people generally use as a ping node? What network do your ping nodes sit on? How people get around the problem that I've described above?

Help please!

[hfourie]

i have a similar situasion. I have 2 nics per san for iscsi traffic and 2 (bond) for lan. What im plannig to do, is assign 1 of the iscsi ports per san a "SAN" ip address which i will use for communication between the boxes for replication. I will place the ping node on this subnet.

Then conf VIP on both iscsi nics on different networks to each other and the "SAN" repl ip.

I do get that if that san nic cable goes it will trigger a failover, but this is how i understand open-e want you to configure it.

anyone else have a better idea?

[enealDC]

I started typing an answer and as I typed, I realized I had additional questions for Open-E regarding when the software *thinks* it needs to failover/failback.
Since Open-E requires a min of two aux interfaces, what role do they play in determining when to failover? What happens if the ping node is down but the auxs are up? What happens if the reverse is true?