Open kinoxyz1 opened 5 days ago
because binpack algorithm. https://volcano.sh/en/docs/plugins/#binpack
because binpack algorithm. https://volcano.sh/en/docs/plugins/#binpack
So if I delete the binpack configuration in volcano schedule conf, can the scheduling be more uniform on each node?
volcano-scheduler.conf: |
actions: "enqueue, allocate, backfill"
tiers:
- plugins:
- name: priority
- name: gang
enablePreemptable: false
- name: conformance
- plugins:
- name: overcommit
- name: drf
enablePreemptable: false
- name: predicates
- name: proportion
- name: nodeorder
- name: binpack ## delete this
yes, you are right
yes, you are right
thank you
Have your problem been solved?
Please describe your problem in detail
my k8s node is as follows:
each node is 16c 64g, the master setting is not schedulable, and the label
node=realtime
is set for the three nodes prod.ds.12.idc、prod.ds.12.idc、prod.ds.14.idcI created a queue:
I try to use scheduled tasks to trigger many Kubernetes Deployments at the same time, so that they can be deployed and run in the realtime queue. however, it was found that the pods were not evenly distributed on the three nodes.
This is the Deployment yaml:
This is the volcano scheduler conf:
This scheduling puts a lot of pressure on two nodes, while some nodes are idle. How can we make the scheduling more balanced?
Any other relevant information
Kubernetes version: v1.21.9 os: CentOS Linux release 7.9.2009 (Core)