My work has 2 San pairs that run Debian Linux (Wheezy) and serve iscsi to a HyperV 2008 R2 cluster using SCST, DRBD, and Corosync via the windows Iscsi. The san nodes are connected to each other via 4GBps link (4 x 1GBps link in LACP) it is never near maxed out. They connection to the hyperVisors (2 x 1 gbps link in LACP) are never maxed either. The cluster runs about 45 VM's (Mix of Windows and Linux) on them. Anywho if any thing IO intensive happens on any of the hosted VM's the disk performance of everything slows down to terrible. In my attempts to find help on this and to repair this issue I found one little snippet in comments section about Windows Iscsi doing wierd things that doesn't bode very will with DBRD back storage targets. However I can't find any more information on this. Has anyone heard anything about Windows Iscsi and DRBD backed storage good or bad?
Thank you
[link][comment]