mirror of
https://github.com/cozystack/cozystack.git
synced 2026-10-26 21:17:32 +09:00
Aleksei Sviridkin
3b375bcf8f
The GPU reconciler always runs, so GPU finalizers and usage records are handled on every cluster, including ones that later turn GPU off; --enable-gpu, which the platform sets only where KubeVirt runs a GPU variant, adds placement, the KubeVirt watches and the webhooks. The webhooks are registered only then because asking the manager for its webhook server starts it, and its certificate is mounted only with GPU enabled. The hoster's resource-name list and raw-name policy come in as flags and reach the admission policies through the settings ConfigMap. A GPU admission waits for the binder only seconds, so every replica must answer from warm caches: the informers the webhooks read start before the manager on every replica rather than at the first admission, readiness waits for exactly those to sync, and the leader releases its lease on shutdown so a rollout does not leave bind requests unanswered for a lease duration. The ConfigMap cache holds only the settings object. Non-positive grace periods are refused, since a zero attachment grace would clear every attachment before its VMI appears. Signed-off-by: Aleksei Sviridkin <f@lex.la> Assisted-by: LLM