Update policy for replacement of managed files

This commit is contained in:
2026-06-18 13:38:44 -05:00
parent 8b0ce4d134
commit 69043801d0
18 changed files with 330 additions and 107 deletions

View File

@@ -205,7 +205,7 @@ Reference: [Operations](operations.md#forced-replacement-workflow).
Symptom: `fail_conflict`, `destination source id differs`, `same id and created time but different digest`, `pipeline id ... does not match`, or `destination id ... does not match`.
Likely cause: `.distributor.json` belongs to a different pipeline, destination, source id, shared-root output owner, or same-created source with different content. Valid managed identity, source, and shared-root output-owner mismatches can publish as `replace_takeover` only when destination `takeover.mode` allows them.
Likely cause: `.distributor.json` belongs to a different pipeline, destination, source id, shared-root output owner, or same-created source with different content. Valid managed identity, source, and shared-root output-owner mismatches can publish as `replace_takeover` when destination `takeover.mode` allows them, or as `replace_conflict` when `transfer.on_conflict: replace` allows managed conflict replacement.
Diagnostic:
@@ -214,7 +214,7 @@ cat <destination-path>/.distributor.json
go run ./cmd/distributor inspect <source-root>
```
Safe fix: verify the source and destination are intended to match. Use a separate destination path for unrelated content. For normal managed replacement, configure destination `takeover.mode` to match the intended ownership boundary. To force exceptional replacement, configure `transfer.on_conflict: replace`, preview with `--dry-run --force`, then publish with `--force`.
Safe fix: verify the source and destination are intended to match. Use a separate destination path for unrelated content. For normal managed replacement, configure destination `takeover.mode` to match the intended ownership boundary or configure `transfer.on_conflict: replace`, then preview with `--dry-run`. Use `--force` only for exceptional replacement of unmanaged content or other force-only cases reported as `force_replace`.
Reference: [Operations](operations.md#destination-state-and-retry-behavior).
@@ -262,7 +262,7 @@ Reference: [Operations](operations.md#managed-output-pruning).
## Destination Is Newer Than Source
Symptom: `skip_destination_newer` or `destination is newer and replacement requires --force`.
Symptom: `skip_destination_newer`.
Likely cause: the destination state records a source manifest with a later `created` timestamp than the current source.
@@ -272,7 +272,7 @@ Diagnostic:
go run ./cmd/distributor run --config <config-path> --dry-run --format json
```
Safe fix: keep the default skip behavior unless replacement is intentional. To replace newer state, configure `transfer.on_destination_newer: replace`, preview with `--dry-run --force`, then publish with `--force`.
Safe fix: keep the default skip behavior unless replacement is intentional. To replace newer valid managed state, configure `transfer.on_destination_newer: replace`, preview with `--dry-run`, then publish without `--force`.
Reference: [Operations](operations.md#forced-replacement-workflow).