2024-11-17T14:23:37.632Z

<Yonatan Zaken> Hi All,
A question relating to ceph containers PidsLimit.

I am using ceph reef 18.2.4
I've noticed that by default, cephadm sets the `--pids-limit=-1`  (unlimited num of pid's) in the unit.run files of all ceph services.
I think last changes in quincy  release were introduced by this PR:
<https://github.com/ceph/ceph/pull/50448/commits/0a37b1cbaee84974188a146247bb08b662904082>

Does anyone know what the default `--pids-limit` value was before quincy release?
And why by default are all ceph containers with unlimited number of pid's?

Thanks 🙂

2024-11-17T14:45:30.678Z

<Matan Breizman> The commit message specifies the default values and the reason for limitless pid:
```The default pids-limit (docker 4096/podman 2048) prevent some
customization from working (http threads on RGW) or limits the number
of luns per iscsi target.```

2024-11-17T14:55:38.764Z

<Yonatan Zaken> So the default pids-limit mentioned in the commit message was the default that was set by ceph before quincy release correct?

Because from `podman run --help`  the default pids-limit is 32768

2024-11-17T15:12:06.976Z

<Matan Breizman> I'm not sure, perhaps the default was changed between version.
It seems like the original PR is based on podman's 3.4.1 introduction to limitless pid with -1 (from the default of 4096):
<https://github.com/containers/podman/pull/12040/commits/14509a92bb87bc328c67182b9e985fd81f6f7b53>

2024-11-17T15:12:18.905Z

<Matan Breizman> I'm not sure, perhaps the default was changed between versions.
It seems like the original PR is based on podman's 3.4.1 introduction to limitless pid with -1 (from the default of 4096):
<https://github.com/containers/podman/pull/12040/commits/14509a92bb87bc328c67182b9e985fd81f6f7b53>

2024-11-17T15:12:24.598Z

<Yonatan Zaken> Thanks Matan

2024-11-17T20:19:16.149Z

<G B> Hello,
I am trying to find out if there is an option to label the nodes and place particular PG for CephFS as per the node labels. Would this work ?
Or will there be a need to modify CRUSH maps ?
If both options are possible, which one would be better ?
My main objective is to do the following :
Create Label for Zone 1 & Zone 2
Add Nodes to Zone1 & Zone2
Make a CephFS pool which has ALL its placement groups in Zone 1
Make CephFS 2 pool which has ALL its PGs in Zone 2
make CephFS3 which has its PGs divided between Zone1 & Zone 2
Any help would be highly appreciated.
PS: I tried using ChatGPT for an answer but it gives me contradictory answers everytime. (edited)