Node failure vs. IO outage

[sandor.bihary]

Dear Support!

I did some tests and examined the failover sequence when I am using 2 node Active-Active StarWind Cluster on Windows2012. The Initiator (2008R2, I need to use that) is using the MS Initiator and MPIO, so the Initiator picked up from both nodes the iSCSI targets.

The test was very simple, I just pulled out the power cable from one of the nodes. After that the Initiator stopped the IO traffic for 30s and after that started to work again.
But when I booted up the "failed" node, the IO stopped again but for 60s!

The first outage I think normal because the iSCSI path have to be failed, but the second outage (which took 2 times longer) when the second node booted up is strange for me.

Could you help me what did I wrong? I don't think this is a normal behavior.

Thanks!
Best Regards
Sanyi

[anton (staff)]

Everything depends on your selected MPIO policy. You should use Round Robin and in this case you'll see no downtime. With failover only you'll see the drops in I/O sequence when turning nodes on and off. See:

http://technet.microsoft.com/en-us/libr ... 51699.aspx

Screenshot of your MPIO config would help.