Hi All
Would like some clarification.
We have a 8 host cluster, which is HA and DRS enabled.
DRS is set to Partially Automated.
My view is if a host fails, DRS will still automatically try to balance the cluster when HA tries to migrate VM's to the remaining hosts.
Is my view correct? or will DRS just make recommendations instead of automatically restarting vm's from the failed host on to the remaining working host.
Any clarification would be appreciated.
Jit
DRS partial mode:
1. Initial placement will be automatic :NO admin monitoring required.
2. Load balancing : Once HA start re-starting VMs on other host (in case of host failure), DRS does monitor cpu/memory load & just provides all possible recommendation, here admin has to monitor all recommendation & apply whichever suits him best. However, while providing recommendation, DRS does consider all the HA constraint. Recommendations would not compromise availability of VMs.
Below documentation will help you to understand how HA works with DRS:
Message was edited by: Vikas :::Added reference for HA DRS interop
Note: Discussion successfully moved from VMware ESXi 5 to Availability: HA & FT
At Partially Automated mode, the initial placement of virtual machines is automatic and based on the performance and resource of hosts... then after a host failure, the VMs will restart on remaining hosts but will not try balance the workload after all virtual machine was restarted... will only make recommendations.
DRS partial mode:
1. Initial placement will be automatic :NO admin monitoring required.
2. Load balancing : Once HA start re-starting VMs on other host (in case of host failure), DRS does monitor cpu/memory load & just provides all possible recommendation, here admin has to monitor all recommendation & apply whichever suits him best. However, while providing recommendation, DRS does consider all the HA constraint. Recommendations would not compromise availability of VMs.
Below documentation will help you to understand how HA works with DRS:
Message was edited by: Vikas :::Added reference for HA DRS interop
if I understand this correct.
HA will still failover VM's to other hosts in the event of a Host Failure.
BUT
DRS will not try to balance the cluster automatically, instead it will just make recommendation.
What happens, if the cluster resources are fragmented amongst the host, would DRS still move VM's around to make best use of resources or will it make recommendations?
Thank you all for your help, much appreciated.
two things here, 1 is HA and 2 is DRS. HA will restart VMs outside of vCenter. DRS does load balancing through vCenter. So VMs restarted by HA will not be placed by DRS, instead HA has its own placement mechanism.
If the cluster is fragmented, the HA will call on DRS to defragment the cluster. This should work either way, regardless of how DRS is configured as far as I understood.
Then again, if your cluster needs to be defragmented... you are doing something wrong! You are using way too many reservations, or are seriously overcommitted. I have never seen this in the real world.
thank you all for your help
So HA will use its own placement mechanism regardless of DRS automation level.
I can not explain this to my manager:smileylaugh:
Correct... We need to, because DRS is a function of vCenter Server. HA cannot rely on vCenter to be available. However. DRS runs every 5 minutes. So if you have it enabled and fully automated any incorrect placement will be correct fast!
