dind - cgroup size

[waw-configit]

I have problem with github runners that are spawning containers using docker in docker.
When dind creates container than process spawned by is not a part of cgroup created for pod (base on pod resource ) i.e.
NAME MemMax
├──kubepods 125 GiB
│ └──burstable
│ ├──arc-runners/dependabot-f6475-runner-tvz4k max
│ │ ├──dind 32 GiB
│ │ ├──runner max
│ │ └──sandbox max

but instead is created under /docker
NAME MemMax
.
├──docker max
│ ├──0673cf40c27800c034e92770c030d231ff7889157b53ec49b6623362812280fd 8 Gib

My questions are:

1. can i force talos, to put process spawned by dind container into cgroup of pod?
2. is there a way to control size of cgroup under /docker, some cgrops under /docker have max as limit.

many thanks!

[smira]

There is nothing that Talos can do here, but my wild guess is that you don't have a proper dind container, as it should still contain itself within a proper cgroup. As you run it as privileged, it can "escape" cgroup tree.

[waw-configit]

@smira Thanks for clarification!