Add Cosmos 1.14.2 row to the watcher-retry behavior history#2742
Open
tatiana wants to merge 2 commits into
Open
Add Cosmos 1.14.2 row to the watcher-retry behavior history#2742tatiana wants to merge 2 commits into
tatiana wants to merge 2 commits into
Conversation
Walk each of the six watcher-retry tables forward by one release, capturing every retry-mechanism PR that ships in 1.14.2: - "Does the Airflow state match dbt's?" — note the upstream-failure recovery fix from #2684. The 1.14.1 row stays as it shipped; the 1.14.2 row carries the closes-the-gap story. - "Task-level retry — consumer" — log the two consumer-side changes: upstream_failure skip-to-fail rewrite from #2684 and the --log-format json strip from #2713 (with the opt-in for users who still want JSON output on retry). - "Task-level retry — producer" — annotate the 1.14.1 known-issue list with "(both fixed in 1.14.2)" and add the 1.14.2 row noting how #2629 (key sanitisation) and #2683 (per-TaskGroup uniqueness) close them. - "Automatic retries" and "Full DAG / TaskGroup clear" — both now cross-reference the producer-row fixes so users don't have to chase down the same two PRs in two places. - "Avoid duplicate or concurrent runs" — note that depends_on_past on a DbtTaskGroup (#2615) now treats the group as an atomic unit across DAG runs, closing the cross-run concurrent-run risk from #2596. 🤖 Generated with Claude Code (https://claude.com/claude-code)
Two consistency / accuracy fixes pointed out in review: 1. "Does the Airflow state match dbt's?" — 1.14.1 row no longer contradicts the new 1.14.2 false-green note. It was previously a bare "Yes", but #2684 (added in 1.14.2) closes a real false-green gap that existed in 1.14.1: upstream-failure-skipped downstream nodes were silently marked SKIPPED while the DAG completed success. The 1.14.1 row now reads "Mostly, except upstream- failure-skipped downstream nodes could produce a 'false green' -- fixed in 1.14.2 by #2684", with the explicit forward pointer. 2. "Avoid duplicate or concurrent runs" — 1.14.2 row corrected to describe the actual topology landed by #2615. Previously read "the producer-done gateway is wired downstream of every consumer task and the task-group leaves carry wait_for_downstream=True", which overstates both halves: only leaf consumer tasks (those without downstream tasks) get wired upstream of the gateway (see cosmos/airflow/graph.py:795-796), and wait_for_downstream=True is set on the producer plus all watcher consumer tasks, not just task group leaves (see graph.py:777, 793). Reworded to: "leaf consumer tasks are wired upstream of the producer-done gateway, and the producer plus watcher tasks carry wait_for_downstream=True." Verified: hatch run docs:sphinx-build -W -b html docs docs/_build/html succeeds. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
tatiana
requested review from
a team,
corsettigyg,
dwreeves and
jbandoro
as code owners
Contributor
There was a problem hiding this comment.
Pull request overview
Updates the watcher retry behavior history documentation to include Cosmos 1.14.2, capturing the retry-related changes shipped in that release across the existing history tables.
Changes:
- Add 1.14.2 rows across the watcher retry history tables, referencing the relevant PRs/issues.
- Update the 1.14.1 “Known issues with the XCom backup mechanism” header to indicate those issues are fixed in 1.14.2.
- Expand narrative details for 1.14.2 around upstream-failure downstream retries and consumer retry logging behavior.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
54
to
+57
| * - **1.14.1** | ||
| - **Yes**. | ||
| - **Mostly**, except upstream-failure-skipped downstream nodes could produce a "false | ||
| green" — fixed in 1.14.2 by | ||
| `#2684 <https://github.com/astronomer/astronomer-cosmos/pull/2684>`_. |
| same DAG run share one backup key and log ``UniqueViolation`` on every model completion. | ||
| * - **1.14.2** | ||
| - Same as 1.14.1, with both XCom-backup known issues from 1.14.1 fixed. The key generator now | ||
| sanitises every component to ``[A-Za-z0-9_-]`` — the safest subset across the secrets |
pankajkoti
approved these changes
Jun 2, 2026
Contributor
pankajkoti
left a comment
pankajkoti
left a comment
There was a problem hiding this comment.
LGTM. Would be nice to alter the PR description. There's a footer line in the PR description which says something like "Drafted, .... " Looks like a status update and some actionable items that are (based on the 1.14.2 release) no longer needed. would be nice to change/remove that too.
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.
Summary
Walk each of the six tables in
docs/guides/run_dbt/airflow-worker/watcher-retry-history.rstforward by one release, capturing every retry-mechanism PR that ships in Cosmos 1.14.2.PR coverage
depends_on_past--log-formatfrom watcher consumer-retry flagsoperator_args={"dbt_cmd_flags": [...]}opt-in pathOne historical-row tweak
In Task-level retry — producer, the 1.14.1 known-issue list header changes from
**Known issues with the XCom backup mechanism:**to**Known issues with the XCom backup mechanism (both fixed in 1.14.2):**. The two listed bullets stay verbatim — readers landing on the 1.14.1 row see the issues as documented, with a forward-pointer to the version that resolves them.The 1.14.1 row in Does the Airflow state match dbt's? is intentionally not edited. It stays as the
**Yes.**it shipped with; the new 1.14.2 row carries the "false green was here, now closed" story so the historical rows remain immutable.Test plan
hatch run docs:buildsucceeds — the 2 build warnings on the resulting build are pre-existing onmain(docs/guides/run_dbt/customization/scheduling.rstduplicate-target labels), unrelated to this change.python3 scripts/check_docs_style.py docs/guides/run_dbt/airflow-worker/watcher-retry-history.rst— clean.codespell docs/guides/run_dbt/airflow-worker/watcher-retry-history.rst— clean.-), and PR-link\#NNNN `_` formatting match the rest of the file.Drafted; promote out of draft once the 1.14.2 stable release goes out (or sooner if you want to merge the doc-update first and let the release-notes link point at it).
🤖 Generated with Claude Code (https://claude.com/claude-code)