Proxmox native CRS vs Proxcenter DRS - do they clash or compliment each other?

[neilcolyn-jpg]

On the latest version of Proxmox 9.2.2, they have improved CRS whereby you can switch HA from "Default (basic)", to "Dynamic Load" and enable "automatic rebalance HA resources" option.

What is the risks/differences between those new settings and Proxcenter's automatic DRS settings?

• Do they perform the same function and if not, how are they different?
• Can we have them both enabled or should we only use one or the other?
• If they both enabled, can it cause issues/clashes or do they compliment each other?

Thanks!

[adminsyspro]

Great question, and the timing is relevant because Proxmox 9.2's "Dynamic Load" scheduler narrows the gap between the two.

Short answer: they overlap, and you should let only one of them automatically rebalance a given guest. They complement each other well if you split the jobs by role rather than running both balancers at once.

Do they do the same thing?

Partly. Both look at node load and live-migrate guests to balance a cluster. The differences that matter:

	Proxmox CRS (9.2 Dynamic Load + auto-rebalance)	ProxCenter DRS
Scope	HA-managed guests only	All guests (VMs and LXC), HA or not
Lives in	The cluster itself (HA manager)	ProxCenter orchestrator, on a schedule
Failover placement	Yes, this is HA's core job	No, it leaves failover to Proxmox HA
Affinity / anti-affinity rules	No	Yes (plus node-affinity)
Modes	On/off	Recommend-only, partial (high-priority only), or fully automatic
Extras	None beyond placement	Scheduling windows, per-cluster concurrency caps, cooldowns, storage-safety checks, fleet-wide view

So CRS is the cluster's built-in placement brain for HA resources, and ProxCenter DRS is a broader, policy-driven balancer that also covers non-HA guests and affinity rules.

Can both be enabled, and do they clash?

Technically yes, but if you turn on "automatic rebalance HA resources" in Proxmox and run ProxCenter DRS in automatic mode, you have two independent loops moving the same HA guests on different schedules with different scoring and no awareness of each other. Each one reacts to the load shift the other just caused, so you can get migration ping-pong: guest moves, the other balancer sees a new imbalance, moves it back. Single moves stay correct, but over time you waste live-migration bandwidth and add churn. One caveat to be aware of: ProxCenter DRS does not currently let you exclude HA-managed guests from automatic balancing (you can scope by cluster or node, or use recommend-only mode), so you cannot cleanly say "Proxmox balances the HA guests, ProxCenter balances the rest."

Recommended setups (pick one):

1. ProxCenter DRS as the single balancer (recommended for most). Leave Proxmox HA Scheduling at Default/basic (or Static) and keep "automatic rebalance HA resources" off. Proxmox HA still does what only it can do, intelligent node selection on failover and recovery, and ProxCenter DRS handles day-to-day balancing for both HA and non-HA guests plus your affinity rules. The two complement cleanly because only one periodic rebalancer is active.

2. Proxmox-native balancing. Enable "Dynamic Load" + "automatic rebalance HA resources", and set ProxCenter DRS to recommend-only (manual) mode, or exclude that cluster from automatic DRS. ProxCenter then stays as visibility and on-demand migrations. Trade-off: you lose automatic balancing for non-HA guests and automatic affinity enforcement.

3. Avoid: both in automatic mode over the same HA guests.

If most of your important guests are HA and you mainly want hands-off balancing with affinity rules and LXC coverage, option 1 is the cleanest. If you want everything handled inside Proxmox and only use ProxCenter for visibility, go with option 2.