6 Commits
Author SHA1 Message Date
Drew Newberry 2538bead50 fix(cluster): pass resolv-conf as kubelet arg and pin k3s image digest (#701) 2026-03-31 09:35:31 -07:00
Piotr Mlocek 0eebbc840f fix(docker): restore apt cleanup chaining in cluster image (#702) 2026-03-31 08:55:31 -07:00
Evan Lezar 122bc74948 feat(sandbox): switch device plugin to CDI injection mode (#503)
* feat(sandbox): switch device plugin to CDI injection mode

Configure the NVIDIA device plugin to use deviceListStrategy=cdi-cri so
that GPU devices are injected via direct CDI device requests in the CRI.
Sandbox pods now only require the nvidia.com/gpu resource request —
runtimeClassName is no longer set on GPU pods.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2026-03-31 08:35:39 -07:00
John T. Myers a831a8921b feat(settings): gateway-to-sandbox runtime settings channel (#474)
* feat(gateway/sandbox): add global and sandbox runtime settings flow
2026-03-20 14:08:57 -07:00
Evan Lezar 495fe4cc73 fix(docker): set migrations dir permissions to 755 on COPY (#475)
Docker COPY preserves build-host file permissions. On hosts with a
restrictive umask (e.g. 0027), the migrations directory is copied as
750 root:root, making it unreadable by the openshell user at runtime
and causing the server to crash with a permission denied error on
startup.

Using --chmod=755 ensures the directory is always world-readable
regardless of the build host umask.

Signed-off-by: Evan Lezar <elezar@nvidia.com>
2026-03-20 14:09:24 +01:00
Drew Newberry a912848217 refactor(build): unify image build graph for cache reuse (#390) 2026-03-18 15:01:04 -07:00