fix(gpu-operator): drop unused hostPID on driver-compat DaemonSet
Signed-off-by: Arsolitt <arsolitt@gmail.com>
This commit is contained in:
parent
2cc60f170c
commit
95ea20119e
1 changed files with 4 additions and 1 deletions
|
|
@ -24,8 +24,11 @@ spec:
|
|||
metadata:
|
||||
labels:
|
||||
app: nvidia-driver-compat
|
||||
# DaemonSet rather than a one-shot Job: staging is idempotent per-node,
|
||||
# must re-run after every Talos reboot (hostPath contents are wiped on
|
||||
# reboot when the system extension re-installs), and the pause container
|
||||
# keeps the pod visible for operator observability (kubectl get pods).
|
||||
spec:
|
||||
hostPID: true
|
||||
priorityClassName: system-node-critical
|
||||
# Restrict to GPU nodes only. Without this the DaemonSet schedules onto
|
||||
# every node (including control-plane and CPU-only workers) and burns
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue