We are running a FAS3070 cluster. On 1 node we have iSCSI luns that our SQL servers point to. On our SQL server monitoring tools we are seeing high disk queues lengths (>5, sometimes above 10 and even above 100). The interesting part is we check the filer, using Performance Monitor, and it's not having any problems. CPU isn't above 50%, sometime only around 25%, Protocol latencies are below 1 millisec, Network Throughput seems fine and Protocol Ops isn't max.
Does anyone know of an issue with external monitoring tools trying to interface with NetApp filers and reporting disk queue lengths? At issue here is can the monitor get a correct number due to the use of multiple disks being used in the aggregate and therefore the lun. Or does the filer head "translate" the multiple disk queue lengths into 1 combined queue length?
Trying to prove to the SQL DBAs that it's not the filers having a problem. lol
As I was typing this, the SQL monitoring tool raised an alarm of an average disk queue length of 5.3. Filer is showing 7% CPU usage, ~20 mb_per_sec network throughput, ~400 iSCSI_ops per sec.
Ideas? And yes it may be the network but trying to prove it's not the filers.
Other way to look at it is what average disk queue length shows the filers are slowing down the process?
Thanks,
Keith
