request and limit settings?

[dashap]

Hello,

For runner scaleset setup (i.e., controller and then runner scalesets with values.yaml files), we are seeing an issue.
A user runs something like yarn and it seems to grab all cpus of the node because it looks at os. It then does not even use all the cpus.

If I run lscpu on a runner pod, I see:

lscpu
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Address sizes: 46 bits physical, 48 bits virtual
Byte Order: Little Endian
CPU(s): 64
On-line CPU(s) list: 0-63

I tried to set in runnerscaleset's values.yaml:

resources:
requests:
cpu: 1000m
memory: 1.0Gi
limits:
cpu: 4000m
memory: 8.0Gi

It still seems to see all the cpu of the node though as seen in the lscpu command.
How do we restrict cpu and memory to a pod so that one user's workflow job does not consume an entire node?

How does request and limits work? If I limit to one cpu and 1 gb of memory as shown above but it needs 4 Gb of memory, does that pod have the ability to get the memory it needs?

[dashap]

Note I ran the following too:

free -mh
total used free shared buff/cache available
Mem: 251Gi 4.8Gi 223Gi 4.0Mi 23Gi 245Gi
Swap: 0B 0B 0B

It seems to show on the pod the total as equal to the whole node's memory and you can see it did use 4.8 Gi even though I set request to 1.0Gi.

[dashap]

fyi:

fyi:
I added to runner scaleset values.yaml:

spec:
containers:
- name: runner
image: our.azurecr.io/actions/actions-runner:2.311.0
command: ["/bin/bash","-c","sudo cp -f /etc/resolv.conf /var/spool/postfix/etc/resolv.conf; sudo /usr/sbin/postfix start;/home/runner/run.sh"]
requests:
cpu: 8000m
memory: 16.0Gi
limits:
cpu: 8000m
memory: 16.0Gi