To Reproduce Steps to reproduce the behavior:

1. Deploy RHOSP 17.1 overcloud with collectd
2. Login to a overcloud node
3. "mem_limit: 512m" is written into collectd.json.

   # cat /var/lib/tripleo-config/container-startup-config/step_3/collectd.json
   
       :   
     "mem_limit": "512m",
       : 

1. This setting is supposed to be applied into the container, but actually it's not.
   Memory limit is not set in the collectd container.

   # podman inspect collectd | jq .[0].HostConfig.Memory
   0 

Expected behavior

Collectd container should have the memory usage limitation:

# podman inspect collectd | jq .[0].HostConfig.Memory
536870912

Bug impact

• This allow collectd to consume extremely huge memory, which results in OOM or unstable behavior on the node

Known workaround

• I was wondering if running "podman update --memory 512m collecd" might work.
  But unfortunately this doesn't work due to [1].
  Even if it worked, this setting won't be persistent. If the container is restarted, the setting is reverted [2]
  Therefore, if OOM occurs once, I think this workaround will be reverted.
  I don't think this can be a good workaround.
  [1] https://issues.redhat.com/browse/RHEL-19137
  [2] https://access.redhat.com/solutions/7058735
• I don't come up with any other workarounds...

Additional context

In RHOSP 16, we introduced the following changes to set the limit of collectd memory usage:

• https://access.redhat.com/solutions/6826441
• https://bugzilla.redhat.com/show_bug.cgi?id=2007255
• https://bugzilla.redhat.com/show_bug.cgi?id=1647057

I think this worked well in RHOSP 16.2.
But it seems that it is not working well in RHOPS 17.1.