Similar to vSphere HA allows VM’s to auto recover from host outages and DRS which is known as dynamic placement on VME is again a cluster setting that live migrates VM between host to sort of balance load
Dynamic Placement on the cluster is enabled in the cluster settings
Under the Options tab on the cluster properties you’ll also find the placement profile.
nder the Options tab on the cluster properties you’ll also find the placement profile.
HA needs a heartbeat datastore. Set it under Infrastructure → Storage → Data Stores: edit the shared datastore (shared-ds-01) and tick Heartbeat Target. Until you do, the cluster warns that it can’t fail VMs over.
Under the Options tab on the cluster properties, we see the placement profile
Each VM also has a Each VM also has a Placement setting, the per-VM DRS override:
- Auto: DRS may move it.
- Pinned: stays on its current host, like a DRS rule that keeps a VM on one host.
- Failover: after HA moves it, it goes back to its original host once that host recovers.
Testing HA
With everything working, we powered off hvm-n2 from Proxmox (qm stop 152) (a hard stop, like pulling the power) and kept pinging centos-1:
Morpheus has picked Host 2 is down
The VM went unreachable, HA restarted it on a surviving host, and it answered again.
rom 172.16.11.1 icmp_seq=232 Destination Host Unreachable
From 172.16.11.1 icmp_seq=233 Destination Host Unreachable
From 172.16.11.1 icmp_seq=234 Destination Host Unreachable
From 172.16.11.1 icmp_seq=235 Destination Host Unreachable
64 bytes from 172.16.20.101: icmp_seq=236 ttl=63 time=419 ms
64 bytes from 172.16.20.101: icmp_seq=237 ttl=63 time=0.509 ms
64 bytes from 172.16.20.101: icmp_seq=238 ttl=63 time=0.351 ms
64 bytes from 172.16.20.101: icmp_seq=239 ttl=63 time=0.507 ms
64 bytes from 172.16.20.101: icmp_seq=240 ttl=63 time=0.514 ms
^C
— 172.16.20.101 ping statistics —
HA works..