PrefectHQ/prefect

prefect

Performance History

Latest Results

docs: send referrers on outbound links and add missing page descriptions (#23728) Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>

main

27 minutes ago

Address review: recheck expiry after the holder check; drop the timed sleep - FileSystemLockManager.release_lock now checks the holder first and only then whether the lock still exists, so an expiry landing between the two checks is tolerated the same as one before them. - The memory-manager expiry test drives _expire_lock synchronously instead of racing the 0.1s threading.Timer against a 0.2s sleep. Co-authored-by: ZCode <zcode@zcode.dev>

Wilson0511:fix/lock-release-after-expiry

2 hours ago

chore(deps): Update github actions (#23699) Co-authored-by: Alex Streed <desertaxle@users.noreply.github.com>

main

2 hours ago

Make release_lock a no-op for expired locks in memory and filesystem lock managers (OSS-8403) Co-authored-by: Alexander Streed <alex.s@prefect.io>

devin/oss-8403-release-expired-lock-noop

2 hours ago

Tolerate release of expired locks in MemoryLockManager and FileSystemLockManager Releasing a lock whose hold timeout has elapsed (or which was released during a task retry) is now a no-op in both managers, matching the documented RedisLockManager.release_lock semantics. A live lock held by a different holder still raises ValueError. Transaction.commit/rollback call release_lock unguarded (including in a finally: block), so an expired lock previously failed a SERIALIZABLE transaction's own teardown and could mask the original exception. Also fixes an undefined manager reference (F821) in TestFileSystemLockManager.test_acquire_lock_when_previously_holder_timed_out. Co-authored-by: ZCode <zcode@zcode.dev>

Wilson0511:fix/lock-release-after-expiry

3 hours ago

fix(hooks): check how a flow hook takes its arguments before calling it Calling `on_crashed` and `on_cancellation` hooks positionally fixed #23714, but as pointed out in review, it broke keyword-only hooks that worked in the runner and handed the wrong objects to hooks that name `flow`, `flow_run` and `state` in a different order. So instead of picking one calling convention, the hook's signature is checked first. If the hook names its parameters `flow`, `flow_run` (or `run`, like the states docs) and `state`, it gets them by name, in any order and keyword-only included. Otherwise it gets them positionally, whatever the names are. If neither fits, the hook isn't called and the log says which signatures work. Nothing is retried after a `TypeError`, so a hook that raises one still runs only once. The engine and the runner now share this rule, which also fixes two quiet problems in `on_failure` and `on_completion`: keyword-only hooks never ran there, and reordered parameters got swapped values. The worker bootstrap keeps its own copy of the rule because it runs with the flow's runtime, which may have an older Prefect. `FlowStateHook` now describes both forms, the `on_*` decorators return the hook's own type, and `docs/v3/concepts/states.mdx` lists the accepted signatures. I ran 23 hook signatures through the four call paths on `main` and on this branch: nothing that worked before receives different objects, and 30 cases that failed or got swapped values now work. The one difference for existing hooks is that a hook taking only `*args, **kwargs` now gets positional arguments in the runner too, the same as in the engine.

CassianoE:fix-hook-keyword-call

3 hours ago

Raise a clear ValueError from generate_slug for fewer than 2 words

herniqeu:slug-word-count

12 hours ago

Stop truncated_to from repeating the whole value for tiny lengths

herniqeu:fix/truncated-to-half-zero

13 hours ago

Latest Branches

Tolerate release of expired locks in MemoryLockManager and FileSystemLockManager#23725

2 hours ago

1f8619e

Wilson0511:fix/lock-release-after-expiry

Make `release_lock` a no-op for expired locks in memory and filesystem lock managers#23726

3 hours ago

5d71a82

devin/oss-8403-release-expired-lock-noop

Accept the same flow state hook signatures in the engine and the runner#23715

3 hours ago

a560afc

CassianoE:fix-hook-keyword-call

Home Terms Privacy Docs