Repository navigation
Release Python SDK 2.19.0: all() member failures no longer block runs - #342
Merged
Merged
Conversation
Release the all() member-failure fix that no longer resends an open group (#338, #341), WorkflowContext.schedule_origin (.github#243, #340) and typed request_id_ambiguous signal/update refusals (.github#216, #322). Retarget the installed retry-cap qualification cell to the 2.19.0 candidate on native Server 3.13.0. Co-Authored-By: Claude Opus 5.5 <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Member
Author
|
Candidate review. Build-only run 44343865375 (attempt 1) built head
🤖 Generated with Claude Code |
|
Member
Author
|
Published, with the post-publish qualification not green.
🤖 Generated with Claude Code |
This was referenced Nov 18, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.



Prepares the Python SDK 2.19.0 release from current
main(88e3075). Not tagged or published.What ships (since 2.18.0)
Package changes since 2.18.0 are #341, #340 and #322. They add public API and a fix with no breaking change, so this is a minor release.
all()member failure while siblings are still running no longer resends the group or gets409 parallel_sequence_mismatch. A failure raised early is recorded as adurable_workflow.parallel_failure_decisionside effect.WorkflowContext.schedule_origin(Let scheduled runs read the schedule occurrence that started them (.github#243) #340, .github#243).request_id_ambiguoussignal/update refusals (Qualify installed Python command-key retries on Server 3.13.0 #322, .github#216).This PR changes the version metadata (
pyproject.toml), the CHANGELOG, and retargets the installed retry-cap qualification cell (ci.yml,activity_acceptance_release.py,published-retry-cap.json, its test and README) from 2.18.0 on Server 3.12.0 to the 2.19.0 candidate on native Server 3.13.0 (indexsha256:30ccaf2c…, source2d1244be, amd64sha256:6cdcb909…).Release notes: the visible effect, checked live
On native Server 3.13.0 (default config, local capped containers), with a 7-member
all()whose second member fails at once, caught and compensated:409 parallel_sequence_mismatch, run stayswaitingrecord_side_effect+ compensation in one task (no resend)waiting. The Server has no workflow task for it and sends no new onewaiting.repairrefusedrepairwhile a member still runsdevelopment_workflow_task_repair_active_work_unqualifiedrepairafter every member settled, no signalrepair_dispatched. The 2.19.0 worker raised the failure, compensated, run completedSo the CHANGELOG states both effects precisely: workflow code after a group that failed early starts one sequence later, and runs blocked on 2.18.0 recover after the upgrade plus a repair once the group's members have finished. Upgrading alone does not wake them, and a blocked run that received a signal or update cannot be repaired on Server 3.13.0.
Local checks
ruff check .passes. The focused suites pass: release metadata, retry-cap cell/workflow/download,test_parallel_member_failure_open_siblingsand the replay regression corpus (209 passed). Onmain, CI fora9d5c76failed once, in Python 3.10test_discard_keeps_generator_context_tokens_and_cannot_borrow_parent_time[cyclic-finalizer](a GC-finalizer timing test). The nextmainrun at88e3075passed the same cell.Remaining release steps
.github/release-candidates/2.19.0.jsonon this branch.gh workflow run ci.yml --ref release/sdk-python-2.19.0 -f activity_retry_cap=true -f retry_cap_artifact_source=candidate.--match-head-commit).2.19.0on the merge commit, then push it. That triggerspublish.yml.ci.ymlonmainwithretry_cap_artifact_source=registry.🤖 Generated with Claude Code