starred/windmill
1
0
Fork 0
mirror of https://github.com/windmill-labs/windmill.git synced 2026-10-13 00:03:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

docs: note the out-of-memory behavior of non-privileged workers

Browse Source 
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Ruben Fiszel
2026-10-06 17:36:05 +02:00
co-authored by Claude Opus 5.5
parent 40d438fe47
commit 62685d7612
1 changed files with 2 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
examples/deploy/nsjail-security-profiles/README.md
+2
View File
@@ -60,6 +60,8 @@ The `DISABLE_NSJAIL=false` environment variable is the fallback: it enables nsja
| Added capabilities | none | `SYS_ADMIN`, `SETPCAP`, and optionally `SYS_RESOURCE` |
| Pod Security Standards | `baseline` on 1.35+; 1.33 and 1.34 reject `procMount: Unmasked` unless the `UserNamespacesPodSecurityStandards` feature gate is on | needs an exemption for the added capabilities |
Neither set of pod settings is privileged, so the worker cannot clear `memory.oom.group` on its own cgroup: from Kubernetes 1.32, a job that exceeds the container's memory limit takes the worker down with it instead of being killed alone. Setting `singleProcessOOMKill: true` in the kubelet configuration of the nodes restores the per-job kill.
## What the profiles give up
Jobs inherit the seccomp filter, so job code can also call `clone` with namespace flags and `mount`, which the runtime default denies to a container without `SYS_ADMIN`. Where the kernel allows unprivileged user namespaces, a job can therefore create a nested user and mount namespace and mount filesystems inside it. This grants nothing outside that namespace, but it exposes more of the kernel to job code than the runtime default does.