- High latency spikes in the datastores during business hours (8 - 5)
- Latency spikes to 1000ms for short (1min) spikes, baseline latency during those periods around 200-300ms. The intervals last for 10 - 30 minutes. See attached image for a visual.
- before and after an "event", the latency baselines under 20ms, with spikes to 30ms.
I have 2 x3650 hosts (1 is an M3, the other an M4)
Each has 2 SAS HBA's connected to each controller on the DS3200
DS3200 - 3 x 300g SAS in RAID5, 3 x 600g SAS in RAID5 (disk #7 global host spare)
- host group, shared access, set to VMware OS.
Hosts are running same build esxi 5.0 or 5.1 (I've tried both).
Total VM's: 10, Windows 2003/2008, Exchange, MS SQL, Terminal (nothing exotic)
If I vMotion all VM's to one host (call it host#1), then power down the other host (host#2), I have no latency spikes.
As soon as I power on the host #2 (boots to internal storage, in maintenance mode), the latency on the VM's (on host #1) will still spike for periods.
Reversing, I can vMotion all VM's to host #2, and power down host #1 - the VM's run fine on host #2. Powering on host #1, I start getting latency spikes.
Looking for assistance...thoughts...help?