Hello, I have a new vsan cluster made up of 3 nodes. As I migrate VMs into the vsan datastore, I am seeing VMs that are already in the vsan datastore have high disk latency numbers. Numbers as high as 500ms.. If I migrate more than 1 VM at a time, the numbers go higher.
Are my expectations set correctly? Storage vmotion events are all write and intensive .. but I have never seen VMs that reside on the same datastore be this affected by a storage vmotion event.
I have support cases open with both Dell and vmware. The one thing they had me do was to downrev the Megaraid driver. 5.5U2 put a 6.X driver on. The HCL showed 5.34 for the the PERC710 on 5.5... (although the HCL for the PERC 710 on the vsphere 6 has the 6.x driver). I think its a non-issue, but I did it. They have been having me run vsan explorer and save the logs and then upload them and total system logs from the cluster.
With 400GB of SSD per raid group -- they should be taking the bulk of the writes for the storage vmotion. Most of the VMs that I am moving into the cluster are under 100GB each.
Any thoughts? Thanks. Paul
Cluster details (each node):
Dell R720, 256GB RAM
Dell PERC H710
10K RPM SAS HD - Seagate ST1200MM0007 -- 1TB (14)
SSD - SanDisk Pliant LB406M -- 400GB (2)
Connected via dual 10Ge.
vDS with port groups for vmk's
there were similar discussion in another thread.
Your SSDs are class C endurance grade
Capacity: 400 GB
Interface Speed: 6 Gbps
Form Factor: 2.5"
Endurance Class: Endurance Class B >=1825 TBW
Performance Class: Class C: 10,000-20,000 writes per second
Flash Technology: SLC SSD
also by default the VM policy is FTT=1 (failure to tolerate). what this translates to is 2 copies of data. so 1 VM migrate of 100GB means 200GB worth of data writes are happening
it will be useful to look at VSAN Observer stats/performance charts to understand whats going ?