fix(helm): start PostgreSQL without gosu on K3s
This commit is contained in:
1 parent
60356c55aa
commit
88ebfd30e1
7 files changed
+124
No files matched your search
@@ -35,6 +35,22 @@ helm template ci-edge release-charts/mixingx-edge-unit
|
||||
|
||||
`mixingx-deploy` 的 ApplicationSet 按 unit chart 逐个发布,每个 edge/agent 一个独立 Helm release。详见 `mixingx-deploy/`。
|
||||
|
||||
## PostgreSQL on K3s
|
||||
|
||||
PostgreSQL StatefulSets use the Debian image's UID/GID 999 and run the image
|
||||
entrypoint directly. This skips gosu while preserving database initialization
|
||||
and startup on the existing data volume. The readiness probe uses
|
||||
`pg_isready` over TCP so the container is Ready only after PostgreSQL accepts
|
||||
connections, including after recovery.
|
||||
|
||||
Only the PostgreSQL container uses `privileged: true` as a temporary workaround
|
||||
for checkpoint signal failures observed on affected Ubuntu 26.04 / K3s nodes.
|
||||
The namespace's admission policy must permit privileged containers. The
|
||||
[containerd signal-denial report](https://github.com/containerd/containerd/issues/14485)
|
||||
describes the same checkpoint error with AppArmor. Check the host kernel log
|
||||
for an AppArmor signal denial and validate initialization and recovery with the
|
||||
corrected runtime or profile before removing privileged mode.
|
||||
|
||||
## 关联文档
|
||||
|
||||
- hub [AGENTS.md](https://gitee.com/mixing-tech/mixingx.hub) — 全项目指南
|
||||
|
||||
Reference in new issue
Block a user