cloudnative-pg / cloudnative-pg/plugin-barman-cloud
`isWALArchiever` is ignored when multiple Barman Cloud plugins defined
- 主要言語
- Go
- スター
- 191
- フォーク
- 72
- 平均マージ
- 1日 16時間
- マージ済み PR(30日)
- 18
説明
## Summary
When defining **multiple entries with the same plugin name** (`barman-cloud.cloudnative-pg.io`) in `Cluster.spec.plugins`, WAL archiving appears to use the **parameters of the last matching entry**.
In practice, `isWALArchiver` can look ignored because WALs are pushed using `barmanObjectName` from the last plugin entry.
## Environment
- CloudNativePG: `1.29.1`
- plugin-barman-cloud: `0.12.0`
- Kubernetes: Kind (local)
## Minimal reproduction
Create two object stores:
- `store-a` (intended WAL target)
- `store-b` (intended backup-only target)
Then use a `Cluster` with duplicate plugin entries:
```yaml
apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
name: pg
spec:
instances: 1
imageName: ghcr.io/cloudnative-pg/postgresql:16
storage:
size: 1Gi
plugins:
- name: barman-cloud.cloudnative-pg.io
isWALArchiver: true
parameters:
barmanObjectName: store-a
- name: barman-cloud.cloudnative-pg.io
parameters:
barmanObjectName: store-b
```
Generate write traffic and inspect object stores.
## Expected behavior
WALs should go to `store-a` because that entry has `isWALArchiver: true`.
## Actual behavior
WALs are pushed to `store-b` (the last plugin entry in `spec.plugins`).
## Why this seems to happen
From source review:
1. CNPG validates "at most one WAL archiver" but does not reject duplicate plugin names.
2. CNPG selects WAL archiver by plugin **name**.
3. `plugin-barman-cloud` config extraction overwrites parameters on each matching plugin name while iterating, so the **last matching entry wins**.
## Relevant source references
- CNPG webhook validation (`at most one WAL archiver`):
- `internal/webhook/v1/cluster_webhook.go` (`validatePluginConfiguration`) [1]
- CNPG WAL archiver plugin name selection:
- `api/v1/cluster_funcs.go` (`GetEnabledWALArchivePluginName`) [2]
- plugin-barman-cloud parameter resolution (`last match wins`):
- `internal/cnpgi/operator/config/config.go` (`NewPlugin`, `NewFromCluster`) [3]
## Conclusion / Questions for maintainers
Could you please confirm whether this is the intended behavior when multiple
`spec.plugins` entries share the same plugin name?
If this behavior is intentional:
- Is it documented somewhere (especially the effective "last matching entry wins"
parameter resolution)?
If this behavior is not intentional:
- Are there plans to fix it (for example by rejecting duplicate plugin names in
validation or by making plugin resolution deterministic and explicit)?
Thanks for your time and for maintaining these projects.
[1] https://github.com/cloudnative-pg/cloudnative-pg/blob/758532a6f4bc148f7f73a769e3abbfe3ffbc710c/internal/webhook/v1/cluster_webhook.go#L2788
[2] https://github.com/cloudnative-pg/cloudnative-pg/blob/758532a6f4bc148f7f73a769e3abbfe3ffbc710c/api/v1/cluster_funcs.go#L1547
[3] https://github.com/cloudnative-pg/plugin-barman-cloud/blob/0bb78879ce8202addf1f0d3bbfbf4485f81a1290/internal/cnpgi/operator/config/config.go#L282
コントリビューションガイド
調査の方向性
internal/cnpgi/operator/config/config.go、特に NewPlugin と NewFromCluster から始め、issue にある重複名の設定を再現してください。その解決方法を、CNPG の検証および WAL-archiver の選択に関する参照と比較してください。重複エントリの動作が明示的に処理され、適切なテストでカバーされるか、意図した動作として文書化されていれば完了です。
索引モデルが issue の本文から書いたものです。
評価
- 技術スタック
- go, kubernetes, postgresql
- 領域
- backend, databases
- issue の種類
- バグ
- 難易度
- 4/5
- 見積もり時間
- 3〜5日
- 活発さ
- 静か
- 明瞭さ
- おおむね明確
- 初心者へのやさしさ
- 48/100