Multipath with ETERNUS DX80 iSCSI

[linuxmc]

Hi,
i have to implement a multipath strategy on an infrastructure with one ETERNUS DX80 iSCSI with 2 module.
At the moment the configuration in production is:

One 1 gigabit SWITCH
One VSphere 5 Cluster (2 server connected to the switch above)
One ETERNUS DX80 iSCSI with 2 module (used with cluster) with 1 LUN - 1 VOLUME and all the traffic on the CM#0 PORT#0

I have a third server with ESXi 5 to make tests about multipath/failover for iSCSI and other traffic, and my idea is to configure server, switch and SAN in this mode:

NIC0 connected to SWITCH0 connected to CM#0 - PORT#1
NIC1 connected to SWITCH1 connected to CM#1 - PORT#1

I read that it is necessary to set Reset Group, but i don't know what it could happen if i do this in production environment.So my questions are:
1. What do i need to setup Reset Group ?
2. If i simulate network failure with my third server to see if multipath work, is there the risk to lose connection between the production cluster and SAN ?
Than you very much.

[me@work]

Did you consider following the advice given in the Server Connection Guide? It outlines best practises and suggestions to configure both the storage device AND your hosts.

The effects of testing network failure are dependant on the network failure that you test, obviously. If you only play with cables between the switch and your test host, then this will probably not disturb your production systems. Provided, of course, your switch is not too dumb, and crash due to unexpected signals...

[linuxmc]

Thank you for the answer. I read the version 2013 of your document, but it doesn't answer to all my questions. I read other documents from Fujitsu website but it isn't clear for me if i can do what i want securely, or if the fact that i yet communicate with my LUN using CM0-P0 prevents me to do connection tests on the same LUN using the remaining ports (CM0-P1, CM1-P0 and CM1-P1). Do you think that i can contact support for this question ?
Thank you,
have a nice day.

[me@work]

I'd say it will probably be ok to contact support, but I'd also like to suggest to be very precise then with what actually you want to test.

I, personally, can't imagine a (real-world) connection test from an independent host that would obstruct the other hosts' connections. If that was the case, then any failing host (that failed in a similar way) connected to the Eternus would kill the traffic for all other hosts...

If you set up your connections as described in the Server Connection Guide (and in VMware's iSCSI Best Practices perhaps), then any failure of the modules involved should not stop communication over the remaining redundant pathes.

If, on the other hand, you're testing interference directly on the Eternus's connection modules, well, then anything could happen. But that's not real world anymore, I think.

[linuxmc]

Ok, thank you for the support