We have an 2 node NW6.5 SP3 cluster with 2 mirrored iSCSI nodes (both on
OES) running as file server, DHCP and ifolder server. When we transfer a
high amount (ex 700MB via FTP) of data the performance of the cluster drops
below acceptable. The cluster seems to freeze all actions for 2-4min, no
responce is given to all connected users.
Software patches are on latest level (exept the N65NSS3a but this seems for
other problems). Also latest driver updates are done. Hardware is a mix of
low cost MSI servers and HP260G4 servers. iSCSI nodes use HP cabinets with
raid ctrl. No fancy Fiber cards are used! Lan is 1Gb copper.
I found a syncrone rise of the "current disk requests" to 1000 during the
performance drop, this is during the actual copy process and an certain
period after the copy process.
At this moment we are still testing with a small group of people but +/-
600 should connect later on.
Is this a configuration setting? Or just wrong hardware?