Fixes#342. Builds docs/spec/features/blob-store-cleanup.md (approved in #357).
A resolved change now keeps its old store's settings. action: "cleanUp" goes through the blob index the way the copy does and deletes a blob from the old store only after reading it back from the current store and checking it (hash for content-named blobs, byte for byte for the rest). Anything missing, damaged, different or unreadable stays in the old store and is reported. dryRun: true previews; a real run needs confirm set to the change's id, and the refusal explains what that costs for older backups. When a run is done the old store's settings go.
It's refused while any node notes its blob store somewhere other than the new store (each node now notes this at start-up and on the minute refresh), when the two stores overlap (shared sharded member, or a probe written to one shows up in the other), after a move back to the old store, or while a plan copies ahead there. It shares the one-run-at-a-time lock with copies, can be stopped, blocks destroy while it runs, and its end is written to the audit log by system:BlobStoreCleanup.
New properties: oldStoreKept, cleanup (with its own problems), and write-only dryRun and confirm.
Tests: model unit tests (refusals, node places, forgetting order), the overlap probe and sharded places, and the blob_store_change_tests system suite gained a cleanup section (preview, confirm, node refusal, stop, missing and damaged kept, second run done), which passes locally with RocksDB. The console, CLI and docs follow in their own PRs.
Fixes #342. Builds `docs/spec/features/blob-store-cleanup.md` (approved in #357).
A resolved change now keeps its old store's settings. `action: "cleanUp"` goes through the blob index the way the copy does and deletes a blob from the old store only after reading it back from the current store and checking it (hash for content-named blobs, byte for byte for the rest). Anything missing, damaged, different or unreadable stays in the old store and is reported. `dryRun: true` previews; a real run needs `confirm` set to the change's id, and the refusal explains what that costs for older backups. When a run is done the old store's settings go.
It's refused while any node notes its blob store somewhere other than the new store (each node now notes this at start-up and on the minute refresh), when the two stores overlap (shared sharded member, or a probe written to one shows up in the other), after a move back to the old store, or while a plan copies ahead there. It shares the one-run-at-a-time lock with copies, can be stopped, blocks destroy while it runs, and its end is written to the audit log by `system:BlobStoreCleanup`.
New properties: `oldStoreKept`, `cleanup` (with its own problems), and write-only `dryRun` and `confirm`.
Tests: model unit tests (refusals, node places, forgetting order), the overlap probe and sharded places, and the `blob_store_change_tests` system suite gained a cleanup section (preview, confirm, node refusal, stop, missing and damaged kept, second run done), which passes locally with RocksDB. The console, CLI and docs follow in their own PRs.
A resolved change keeps its old store's settings, and action cleanUp
deletes from the old store each blob the current one has, read back and
checked. What fails the check stays and is reported. dryRun previews;
deleting needs confirm set to the change's id. Refused while a node is
on the old store (each node now notes where its blob store is), when the
stores overlap, after a move back, or while a plan copies ahead there.
Runs one at a time with the copies, and its end is audited.
Spec: docs/spec/features/blob-store-cleanup.md (#342).
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Fixes #342. Builds
docs/spec/features/blob-store-cleanup.md(approved in #357).A resolved change now keeps its old store's settings.
action: "cleanUp"goes through the blob index the way the copy does and deletes a blob from the old store only after reading it back from the current store and checking it (hash for content-named blobs, byte for byte for the rest). Anything missing, damaged, different or unreadable stays in the old store and is reported.dryRun: truepreviews; a real run needsconfirmset to the change's id, and the refusal explains what that costs for older backups. When a run is done the old store's settings go.It's refused while any node notes its blob store somewhere other than the new store (each node now notes this at start-up and on the minute refresh), when the two stores overlap (shared sharded member, or a probe written to one shows up in the other), after a move back to the old store, or while a plan copies ahead there. It shares the one-run-at-a-time lock with copies, can be stopped, blocks destroy while it runs, and its end is written to the audit log by
system:BlobStoreCleanup.New properties:
oldStoreKept,cleanup(with its own problems), and write-onlydryRunandconfirm.Tests: model unit tests (refusals, node places, forgetting order), the overlap probe and sharded places, and the
blob_store_change_testssystem suite gained a cleanup section (preview, confirm, node refusal, stop, missing and damaged kept, second run done), which passes locally with RocksDB. The console, CLI and docs follow in their own PRs.