[SOLVED] VM migraters without reason at startup (HA member)

B

BenediktS

Member
Proxmox Subscriber
Dec 6, 2022
55
19
13
Hello,

we have a HA cluster with 4 Nodes. (PVE 8.2.3)
50% of the VMs that a running in that cluster are running on a HA configuration.
All the HA VMs have "restart: 1, relocate: 1, Groupe: <none>".
They are in no Groupe so no server should be preferred .

Node 1,2&3 are at 80% RAM capacity.
Node 4 is at 20% RAM capacity.

We have a server VM that is turned of and is on Node 4.
When we now start the server VM it migrates to Node 1,2&3 and boots up.
It fills up the RAM and crashes.

1) i would understand the migration, if we had defined a HA group that preferred some Nodes.
2) i might understand the migration if there is a Node with much more free RAM.

But migrating it away from a Node with the most free RAM sounds absolutly wrong.
It even looks like it migrates it just to migrate it. .

I know i could set "Max Relocation : 0" .
But if the server is down i want it to migrate.
I don't want it to migrate if the server is up that it is currently on.

And if it is migrating choose the Node with the lowest RAM usage.
Or at least look if the free RAM on the Node is bigger than the allocated RAM in the VM configuration.

But the best user expreince is, don't migrate it if it is not required.

So my querstion is: Have i overlooked a configration parameter that defines this behaviour ?

best regards
Benedikt
 
i found the solution 2 minutes after i wrote this thread.

DataCenter -> Options -> "Cluster Resource Scheduling"

Choose "Static load" without "rebalance on start".


So the default options is doing rebalancing at the start, even it the checkbox is not checked.
 
You must log in or register to reply here.
Top Bottom