Guys,
We have some brand new 3par Hybrid 8200 hybrid SANs, and some 24 Core Gen10 DL360 servers each with 4 x10gb ports.
Everything is reasonable, aside from snapshots with memory, where it takes 17 mins to snapshot an IDLE Vm with about 4gb of memory.
This is at least twice as slow as our decaying old infrastructure on 1gbe everywhere.
I can sit and watch the esxi hosts feebly farm out bytes over the iscsi vNICs at about 20MB/s sometimes lower. It is absolutely pathetic and infuriating.
After some googling, I saw this article:-
https://virtuallyhyper.com/2013/04/snapshots-take-a-long-time-when-keep-memory-is-checked/
I proceeded to change these totally undocumented advanced settings, and wow, changing to 1ms and 4096 the snapshots are taking 20seconds down from 17 minutes, and I see much better bandwidth usage on the iscsi nics.
Can ANYONE explain what is going on? We have not paid hundreds of thousands for kit to have a 17 mins snapshot time on test VMs doing ZERO work.
If the above settings are "ok" to use, how can I make a change globally to the hosts, so they use this value by default, instead of having to manually change any new VM that is created to add these settings?
I am so confused right now as to how the snapshotting with memory can be so poor over 10gbe network, to an SSD tiered brand new SAN, it is infuriating!
