Ping node/failover and redundant switches.

I'm a little confused on having the ping node setup and what happens with the ping node goes down.

My setup is 2 DSS machines identical setup, 2 dedicated iSCSI only switches. 2 dedicated internet traffic switches.
SAN1 primary node
SAN2 secondary node

Switch01
Switch02
Switch03
Switch04

Switch01 and Switch02 carry all "live/internet traffic"
My router has a bond connection to Switch01 and Switch02, my router is also the ping node. 202.xxx.xxx.1

6 nic's in both dss machines.

eth0 through eth3 is VIP for iSCSI failover and MPIO.
eth0 and eth1 are connected to Switch03
eth2 and eth3 are connected to Switch04
eth5 is directly connected to the other SAN for dedicated volume replication
eth4 on SAN1 is connected to swtich01/eth4 on SAN2 is connected to switch02
eth4 and eth5 are ticked as my auxiliary connections for failover.

Switch03 and Switch04 have a 4 port Trunk with LACP between both of them, they have no "external" connectivity, they can't see the ping node.
Switch01 and Switch02 have a 2 port Trunk with LACP between both of them.

eth0 10.1.5.100
eth1 10.1.6.100
eth2 10.1.7.100
eth3 10.1.8.100
eth4 202.xxx.xxx.70 (.71 on SAN2)
eth5 172.16.1.1 (.2 on SAN2)

I have my ping node as 202.xxx.xxx.1 which is the IP on my core router.

My failover is setup as broadcast.

Soooooo now my questions.
Router is rebooted.
What happens to both nodes? They both will lose contact with the ping node, but my system is still 100% stable.
Switch01 fails, SAN1 will lose contact to the ping node. What happens then? my san network is 100% fine as I have multiple switches and dedicated iSCSI switches, will it failover? likewise if Switch02 fails, what will happen to SAN2?

How do I change my config so that the ping node isn't factored into failover, if I have duel aux connections I don't want the ping node to cause issues.

Cheers
Ad

does this help:

http://blog.open-e.com/ping-node-%e2%80%93-explained/

Excellent, now I get it...

BUT..

You can indicate up to 10 ping nodes. Please use comma as the delimiter (e.g. 192.168.0.1;192.168.0.2 etc.). Automatic failover will only take place if all ping nodes become unavailable to the primary node (e.g. the server does not take action when at least one ping node is reachable).

Is that a COMMA or a SEMI-COLON (thats the help popup on the iSCSI Failover section)?

So in my case could I set 4 ping nodes, being each of the switches01 through 04, that would allow restarts of the swithes and if one of the SANS did die for good, the failover aux links would pic that up?

My next question I gather I have to turn off iscsi failover on my secondary, update the ping node settings turn it back on, failover and then repeat on the primary right?

Cheers
Adam

ugh that won't work actually, all the switches are in the same subnet 202.xxx.xxx.xxx

how do I go about picking my ping nodes then?

I'm think switch01 and switch02 in the 202 subnet, but how do I handle the ping checks on the iscsi network, Im assuming I shouldn't use each san as a ping node? or is that ok to do?

Use a Ping node with UPS, Firewalls, Routers anything that is in the same subnets is also best and has high up time - see link below for more information. we spent allot of time on these video's that will help w/ some of your questions.

http://www.open-e.com/library/webcasts-and-videos/
2010.02 Open-E DSS V6 AutoFailover with VMware 4.0
2010.09 How to setup iSCSI Failover with Unicast Network connection mode

Hi guys,

I've run into a bit of an issue, I'm just not sure how to do this.
As I stated my VIP's are
eth0 10.1.5.100
eth1 10.1.6.100
eth2 10.1.7.100
eth3 10.1.8.100

Which are only active on the active/primary node.
My VMWare hosts have all the iscsi traffic on the network 10.1.5.x/ 10.1.6.x/ 10.1.7.x 10.1.8.x I was going to add the IP's that are on my virtual hosts as ping node IP's but then relised that only the primary server will see them. As the secondary node doesn't have an IP on it in the subnet. This is all a catch 22, none of my routers/firewalls are in the same subnet as my iSCSI traffic, do you guys have any ideas on how I can do this?

If I was to add an ip from one of the above subnets to my switches etc, I'm still have the issue of only the node with the VIP's on it will see the ping nodes.

Any ideas?

Cheers
Ad

If I remember you can add more then 1 ping node and this can be on a different subnet so you should be ok just make sure the ping nodes are seen for each network for the Pri and Sec. Ping them first to be sure for each subnet.