UN-3695 [FIX] PG consumer — supervisor shutdown grace = VT, single shared join deadline by muhammad-ali-e · Pull Request #2153 · Zipstack/unstract

muhammad-ali-e · 2026-07-08T06:07:16Z

What

The PG-queue prefork supervisor SIGKILLed its own children after a hardcoded 30s on shutdown (_SHUTDOWN_GRACE_SECONDS = 30.0), even though the pod's terminationGracePeriodSeconds is set to VT + buffer (9120s for worker-pg-file-processing, VT=9060). So on a graceful SIGTERM (deploy / HPA pod scale-down) the supervisor knifed an in-flight batch ~300× sooner than the grace the pod actually had — orphaning the batch: the executor finished the work, but nothing was left to consume the reply, mark the file COMPLETE, decrement the barrier, or ack the queue message.

Two changes:

shutdown_grace_from_env() — the per-child drain grace now defaults to the consumer VT (WORKER_PG_QUEUE_CONSUMER_VT_SECONDS), floored at 30 so it can never be shorter than before (non-regressive: OSS/dev on the default VT stays 30; fileproc gets 9060 with no chart change, since the chart already sets VT). Explicit WORKER_PG_QUEUE_CONSUMER_SHUTDOWN_GRACE_SECONDS overrides.
_join_children → single shared deadline (was per-child recomputed). Children are SIGTERM'd together and drain in parallel, so one shared window still gives each its full grace while bounding total drain to ~grace. A per-child deadline sums to N×grace and, at grace=VT, blows past the pod grace → k8s SIGKILLs the whole pod, hard-killing siblings that were draining cleanly.

Why

Deploys and autoscaling routinely send graceful SIGTERM to workers. The 30s app-grace silently defeated the ~2.5h drain budget the chart deliberately provisions, so every scale-down/rollout landing mid-extraction stranded a batch until visibility-timeout redelivery (~2.5h) or the reaper.

Scope / safety

The supervisor (pg_queue_consumer) is a PG-worker-only component — it never runs for Celery workers, so no feature-flag branch is needed and the Celery path is byte-for-byte unchanged.
max(30, VT) floor guarantees the drain is never shorter than today.

How tested

Unit: shutdown_grace_from_env (VT-tracking / floor / override) + a shared-deadline regression test — 38 supervisor tests pass; test_pg_consumer_supervisor.py + test_pg_queue_consumer.py = 109 pass.
e2e with real forked children: (1) grace tracks VT (40) and the supervisor drains a live child with no premature SIGKILL; (2) two wedged children are SIGKILLed together at the shared deadline, not 2× it.

Part of UN-3695 (PG crash-safe worker shutdown). Follow-ups tracked there: heartbeat-lease for hard-kill recovery (next); the idempotency track was found already handled (billing is idempotent in the cloud plugins) and needs no code.

🤖 Generated with Claude Code

…ared join deadline The prefork supervisor SIGKILLed its own children after a hardcoded 30s on shutdown (`_SHUTDOWN_GRACE_SECONDS = 30.0`). But the chart grants the pod `terminationGracePeriodSeconds = VT + buffer` (9120s for worker-pg-file-processing, VT=9060) — so on a graceful SIGTERM (deploy / HPA pod scale-down) the supervisor was knifing an in-flight batch ~300x sooner than the pod grace it lives inside, orphaning the batch (executor finished, but nothing consumed the reply / marked COMPLETE / decremented the barrier / acked the message). - `shutdown_grace_from_env()`: the per-child drain grace now defaults to the consumer VT (`WORKER_PG_QUEUE_CONSUMER_VT_SECONDS`), floored at 30 so it can never be SHORTER than before (non-regressive; OSS/dev with the default VT stays 30, fileproc gets 9060 with no chart change). Explicit `WORKER_PG_QUEUE_CONSUMER_SHUTDOWN_GRACE_SECONDS` overrides. - `_join_children` now uses a SINGLE shared deadline instead of a per-child recomputed one. Children are SIGTERM'd together and drain in parallel, so one shared window still gives each its full grace while bounding total drain to ~grace. A per-child deadline would sum to N x grace and, at grace=VT, blow past the pod grace so k8s SIGKILLs the whole pod, hard-killing siblings that were draining cleanly. PG-worker-component only — the supervisor never runs for Celery workers, so no flag branch is needed and the Celery path is untouched. Tests: shutdown_grace_from_env (VT-tracking / floor / override) + a shared-deadline regression test (38 supervisor tests). Verified e2e with real forked children: grace tracks VT and drains a live child (no premature SIGKILL); 2 wedged children SIGKILL together at the shared deadline, not 2x. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

coderabbitai · 2026-07-08T06:07:23Z

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 58170876-7652-4b1f-b26b-b5b791441212

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests
Commit unit tests in branch fix/UN-3695-supervisor-grace-vt

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

Add a co-located README as a lookup dictionary for the PG-queue transport: config env vars (consumer / reaper / orchestration / routing) with defaults, a concepts glossary (visibility timeout, claim, redelivery, reaper, barrier, poison/re-park, prefork supervisor, shutdown grace, request-reply/callback, last_progress_at, leader election, fairness), and the table catalogue. Complements the 9e-design.md design doc. Docs only — no code change. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

muhammad-ali-e

Automated PR review (PR Review Toolkit: code-reviewer, silent-failure-hunter, type-design, comment, test, simplifier). The core fix is correct — grace defaults to VT floored at 30s, override honoured as-is, and the single shared deadline correctly bounds total drain to ~grace instead of N×grace. Findings below are ranked; the two [med] items are the ones worth addressing before merge. Everything else is low/nit.

muhammad-ali-e · 2026-07-08T06:35:24Z

+    """
+    from queue_backend.pg_queue.consumer import _DEFAULT_VT_SECONDS, consumer_env
+
+    override = consumer_env("SHUTDOWN_GRACE_SECONDS", None, float)


[med] SHUTDOWN_GRACE_SECONDS override is unvalidated — a bad value silently produces the exact failure UN-3695 fixes.

consumer_env(..., float) accepts negative and non-finite values, and max(0.0, override) then collapses them:

SHUTDOWN_GRACE_SECONDS=-30 → max(0.0, -30)=0.0 → deadline now+0 → every child SIGKILLed with zero drain on the next scale-down (the orphaned-batch bug this PR exists to prevent), with no log.

nan → max(0.0, nan)=0.0 → same immediate kill.

inf/1e999 → max(0.0, inf)=inf → _join_children deadline is now+inf → shutdown hangs forever until k8s hard-kills the whole pod.

A negative/non-finite drain budget is never intentional. Sibling concurrency_from_env() already raises on n < 1 — mirror that here:

override: float | None = consumer_env("SHUTDOWN_GRACE_SECONDS", None, float) if override is not None: if not math.isfinite(override) or override < 0: raise ValueError( f"WORKER_PG_QUEUE_CONSUMER_SHUTDOWN_GRACE_SECONDS must be a finite " f"number >= 0, got {override!r}" ) return override

(The override: annotation also restores the X | None intent-declaration convention this module documents at consumer.py:858.)

muhammad-ali-e · 2026-07-08T06:35:24Z

+    deadline = time.monotonic() + grace_seconds
    for slot, pid in fleet.alive_items():
-        if _wait_for_exit(pid, time.monotonic() + grace_seconds):
+        if _wait_for_exit(pid, deadline):


[med] New regression: a clean-exited child iterated after the shared deadline elapses is spuriously SIGKILLed and logged as "did not drain".

With one wedged child consuming the whole shared window, every subsequent child hits _wait_for_exit(pid, deadline) with deadline already in the past. _wait_for_exit's loop is while time.monotonic() < deadline: — so it returns False without ever calling waitpid, even for a child that exited cleanly at T+5s. _join_children then SIGKILLs it and logs "did not drain within the shared %ss grace".

The SIGKILL to the (already-zombie) child is harmless, but the WARNING is a false "batch was hard-killed" alarm — ironic for the PR whose whole point is to stop hard-killing in-flight batches, and it can trip exactly that alert. This is new: the old per-child code recomputed time.monotonic() + grace per iteration, so the deadline was never already-elapsed at entry.

Fix in _wait_for_exit — poll waitpid at least once regardless of the deadline:

def _wait_for_exit(pid: int, deadline: float) -> bool: while True: try: reaped, _status = os.waitpid(pid, os.WNOHANG) except ChildProcessError: return True if reaped != 0: return True if time.monotonic() >= deadline: return False time.sleep(0.1)

muhammad-ali-e · 2026-07-08T06:35:24Z

    fleet = _Fleet(concurrency)
+    grace_seconds = shutdown_grace_from_env()
+    logger.info(
+        "PG-queue consumer supervisor: shutdown drain grace = %.0fs/child", grace_seconds


[low] Startup log says %.0fs/child but the drain is now a single shared total, not per-child.

After this PR grace_seconds bounds the total drain to ~grace across all children (see the _join_children docstring, which explicitly rejects the per-child model). An operator reading grace = 9060s/child with concurrency=4 will mis-model the total window as 4×9060s — the precise per-child-vs-shared misconception the rework fixes. Drop the /child:

"PG-queue consumer supervisor: shutdown drain grace = %.0fs (shared across children)", grace_seconds

muhammad-ali-e · 2026-07-08T06:35:24Z



+def shutdown_grace_from_env() -> float:
+    """Per-child graceful-drain budget (seconds) on shutdown, before SIGKILL.


[low] "Per-child" wording left stale after the switch to a shared deadline (3 sites).

This docstring's first line still calls the value a "Per-child graceful-drain budget", and so does the constant comment at line 66 and the README table row at README.md:25 — yet _join_children now applies it as a single shared total and its own docstring explicitly warns that a per-child deadline would be a bug. Same knob described two contradictory ways within one PR. Suggest a find-and-replace to "graceful-drain budget, shared across all children on shutdown" at all three sites (the _join_children docstring and the SIGKILL warning at line 445 are already correct — align the rest to them).

muhammad-ali-e · 2026-07-08T06:35:24Z

+    if override is not None:
+        return max(0.0, override)
+    vt = consumer_env("VT_SECONDS", _DEFAULT_VT_SECONDS, int)
+    return max(float(_DEFAULT_SHUTDOWN_GRACE_SECONDS), float(vt))


[nit] Redundant float() on an already-float constant. _DEFAULT_SHUTDOWN_GRACE_SECONDS is 30.0 (line 71), so float(...) around it is a no-op. float(vt) is the meaningful cast (vt is an int):

return max(_DEFAULT_SHUTDOWN_GRACE_SECONDS, float(vt))

muhammad-ali-e · 2026-07-08T06:35:24Z

+|---|---|---|
+| `CONCURRENCY` | Prefork consumer children per pod (1 = plain single process) | `1` |
+| `VT_SECONDS` | **Visibility timeout** — how long a claimed message stays hidden before it can redeliver | `30` (chart: `9060` for file-processing) |
+| `SHUTDOWN_GRACE_SECONDS` | Per-child graceful-drain budget on SIGTERM before SIGKILL | `= VT`, floored at `30` |


[low] This row contradicts the glossary in the same file. It calls SHUTDOWN_GRACE_SECONDS a "Per-child graceful-drain budget", but the "Shutdown grace" glossary entry at line 87 correctly says the supervisor drains "using a single shared deadline." Reword to Graceful-drain budget (shared across all children) on SIGTERM before SIGKILL so the two agree.

muhammad-ali-e · 2026-07-08T06:35:24Z

        kill.assert_called_once_with(111, _signal.SIGKILL)
        waitpid.assert_called_once_with(111, 0)

+    def test_shared_deadline_not_per_child(self):


[test] Coverage gaps around the change's load-bearing paths.

Multi-child straggler on the shared deadline is untested. test_shared_deadline_not_per_child only exercises the clean-exit path (_capture always returns True); test_straggler_is_sigkilled_and_reaped uses a single child. A regression re-introducing a per-child deadline inside the SIGKILL loop would pass both. Add a 3-child, all-False (none drain) case asserting one shared deadline and kill/waitpid called 3×.

This assertion is probabilistic. len(set(deadlines)) == 1 only distinguishes shared-vs-per-child if time.monotonic() ticks between iterations; with _wait_for_exit patched to return instantly a per-child bug could yield 3 identical values on a coarse clock and pass. Harden by patching time.monotonic with an advancing side_effect=[100.0, 200.0, 300.0, 400.0] and asserting deadlines == [105.0, 105.0, 105.0].

Negative override branch untested — add SHUTDOWN_GRACE_SECONDS=-5 (pairs with the validation fix suggested on supervisor.py:118).

Malformed env is untested — assert shutdown_grace_from_env() raises ValueError matching the offending var name for SHUTDOWN_GRACE_SECONDS=abc and VT_SECONDS=9060s, pinning the fail-fast-at-startup contract.

muhammad-ali-e · 2026-07-08T06:44:58Z

Thanks — all 7 addressed in 67f7d6e83:

[med] supervisor.py:118 unvalidated override — shutdown_grace_from_env now rejects a negative / non-finite SHUTDOWN_GRACE_SECONDS (math.isfinite + >= 0) and raises with the offending var name, mirroring concurrency_from_env. No more -30/nan → 0 (immediate SIGKILL) or inf → hang. Added the override: float | None annotation.
[med] supervisor.py:442 _wait_for_exit false-alarm — it now polls waitpid at least once regardless of the deadline (while True + explicit exits), so a child iterated after the shared window elapsed isn't spuriously reported "did not drain" / SIGKILLed.
[low] :366 log — now shutdown drain grace = %.0fs (shared across children) (dropped /child).
[low] "Per-child" wording (3 sites) — corrected the shutdown_grace_from_env docstring, the constant comment (line 66), and README.md to "shared across all children", matching the _join_children docstring.
[nit] :122 redundant float() — dropped; now max(_DEFAULT_SHUTDOWN_GRACE_SECONDS, float(vt)).
[low] README.md:25 — row reworded to agree with the glossary ("shared across all children").
[test] coverage — (1) shared-deadline test hardened with an advancing time.monotonic (asserts exact 105/105/105, no longer probabilistic); (2) new 3-child all-wedged straggler case asserting kill/waitpid ×3 and one shared deadline; (3)+(4) negative / non-finite / malformed override + malformed VT_SECONDS=9060s fail-fast tests. Replaced the … or True lambdas with named helpers. 43 tests pass; e2e re-verified (drain + wedge).

SonarCloud's "C Reliability on New Code" should clear on this commit — it was the same override/hang class + the lambda idiom, both now fixed.

…adline false-alarm, docs Review round on the grace=VT / shared-deadline change: - shutdown_grace_from_env: reject a negative / non-finite SHUTDOWN_GRACE_SECONDS override (was silently collapsed by max(0.0, ...) -> 0 = immediate SIGKILL, or inf = shutdown hang) — fail fast like concurrency_from_env; annotate override: float|None; drop the redundant float() on the already-float default. - _wait_for_exit: poll waitpid at least once regardless of the deadline. On the single SHARED deadline a child iterated after the window elapsed was reported 'did not drain' and SIGKILLed — a false hard-kill alarm — even though it exited. - Wording: the grace is a single shared total, not per-child — corrected the docstring, constant comment, startup log ('shared across children'), and the README config row to match the _join_children docstring. - Tests: shared-deadline hardened with an advancing time.monotonic (exact 105/105/105); a 3-child all-wedged straggler case (kill+waitpid x3, one shared deadline); negative / non-finite / malformed override + malformed VT fail-fast; named helpers replace the lambda idiom; float asserts use pytest.approx (Sonar S1244). 43 tests pass; e2e re-verified. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…ervisor-grace-vt

greptile-apps · 2026-07-08T06:57:09Z

Greptile Summary

This PR fixes a critical shutdown bug in the PG-queue prefork supervisor where a hardcoded 30-second grace period caused in-flight batches to be SIGKILLed ~300× sooner than the pod's configured terminationGracePeriodSeconds (~2.5h for file-processing workers), orphaning batches on every deploy or HPA scale-down.

shutdown_grace_from_env() replaces the _SHUTDOWN_GRACE_SECONDS = 30.0 constant with a function that reads the consumer's visibility timeout (WORKER_PG_QUEUE_CONSUMER_VT_SECONDS), floors it at 30s, and respects an explicit SHUTDOWN_GRACE_SECONDS override — all validated to be finite and non-negative.
_join_children shared deadline replaces per-child deadline recomputation with a single shared deadline = time.monotonic() + grace_seconds set once before iterating children; _wait_for_exit is updated to poll at least once even after the deadline to avoid false SIGKILL for children that finished during a sibling's wait. A new README documents the queue's configuration and glossary.

Confidence Score: 5/5

Safe to merge — the change is scoped entirely to the PG-queue prefork supervisor, the Celery path is byte-for-byte unchanged, and the fix is non-regressive (VT floored at 30s).

Both changes are mechanically straightforward: replacing a module-level constant with an env-read function and computing one shared deadline instead of N per-child deadlines. Validation in shutdown_grace_from_env covers negative, non-finite, and malformed values. The at-least-once poll in _wait_for_exit correctly handles the shared-deadline edge case where a later-iterated child finished during a sibling's wait window. The 38 existing supervisor tests plus 11 new tests all target the changed code paths directly.

No files require special attention.

Important Files Changed

Filename	Overview
workers/pg_queue_consumer/supervisor.py	Core fix: adds shutdown_grace_from_env() (VT-tracking, validated, floored at 30s) and converts _join_children to a single shared deadline with an at-least-once poll guard in _wait_for_exit; logic is sound and well-guarded.
workers/tests/test_pg_consumer_supervisor.py	Adds 9 new TestShutdownGraceFromEnv cases (VT tracking, floor, override, validation, malformed inputs) and 2 TestJoinChildren regression tests verifying the shared-deadline property; coverage is thorough.
workers/queue_backend/pg_queue/README.md	New reference doc listing all env vars, glossary terms, and table schemas for the PG queue; purely documentation with no functional impact.

Sequence Diagram

%%{init: {'theme': 'neutral'}}%%
sequenceDiagram
    participant k8s as k8s/HPA
    participant sup as Supervisor
    participant c1 as Child 1 (in-flight)
    participant c2 as Child 2 (idle)

    Note over sup: startup: grace = max(30, VT) e.g. 9060s

    k8s->>sup: SIGTERM
    sup->>sup: _on_term: stopping.set()
    sup->>c1: SIGTERM at T0
    sup->>c2: SIGTERM at T0
    c2-->>sup: exits quickly

    Note over sup: finally block runs
    sup->>c1: SIGTERM again (harmless duplicate)
    sup->>c2: SIGTERM again (ProcessLookupError suppressed)

    Note over sup: _join_children: deadline = now + grace_seconds (ONE shared value)

    sup->>c1: _wait_for_exit(pid1, deadline) polls every 0.1s
    c1-->>sup: exits within grace period
    sup->>c2: _wait_for_exit(pid2, deadline) polls at least once
    Note over c2: already exited, first poll returns True

    Note over sup: all children drained, supervisor exits cleanly

%%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%%
sequenceDiagram
    participant k8s as k8s/HPA
    participant sup as Supervisor
    participant c1 as Child 1 (in-flight)
    participant c2 as Child 2 (idle)

    Note over sup: startup: grace = max(30, VT) e.g. 9060s

    k8s->>sup: SIGTERM
    sup->>sup: _on_term: stopping.set()
    sup->>c1: SIGTERM at T0
    sup->>c2: SIGTERM at T0
    c2-->>sup: exits quickly

    Note over sup: finally block runs
    sup->>c1: SIGTERM again (harmless duplicate)
    sup->>c2: SIGTERM again (ProcessLookupError suppressed)

    Note over sup: _join_children: deadline = now + grace_seconds (ONE shared value)

    sup->>c1: _wait_for_exit(pid1, deadline) polls every 0.1s
    c1-->>sup: exits within grace period
    sup->>c2: _wait_for_exit(pid2, deadline) polls at least once
    Note over c2: already exited, first poll returns True

    Note over sup: all children drained, supervisor exits cleanly

_{Reviews (2): Last reviewed commit: "UN-3695 [FIX] SIGKILL warning log uses %..." | Re-trigger Greptile}

muhammad-ali-e · 2026-07-08T06:59:41Z

Addressed the Greptile nit (20462e044 → amended): the SIGKILL warning in _join_children now uses %.0fs for grace_seconds, matching the startup log (was %s → printed 9060.0s vs the startup log's 9060s). Cosmetic only, no behavior change.

… (Greptile nit) The _join_children SIGKILL warning printed grace_seconds with %s (e.g. 9060.0s) while the startup log uses %.0f (9060s). Align to %.0fs. Cosmetic — no behaviour. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

sonarqubecloud · 2026-07-08T07:03:12Z

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

muhammad-ali-e force-pushed the fix/UN-3695-supervisor-grace-vt branch from 2562182 to e61708f Compare July 8, 2026 06:31

muhammad-ali-e force-pushed the fix/UN-3695-supervisor-grace-vt branch from e61708f to baa0199 Compare July 8, 2026 06:34

muhammad-ali-e commented Jul 8, 2026

View reviewed changes

muhammad-ali-e force-pushed the fix/UN-3695-supervisor-grace-vt branch from 67f7d6e to 20462e0 Compare July 8, 2026 06:48

Merge branch 'feat/UN-3445-pg-queue-integration' into fix/UN-3695-sup…

b38f8b3

…ervisor-grace-vt

muhammad-ali-e marked this pull request as ready for review July 8, 2026 06:50

muhammad-ali-e force-pushed the fix/UN-3695-supervisor-grace-vt branch from b38f8b3 to 56386bc Compare July 8, 2026 07:00

muhammad-ali-e force-pushed the fix/UN-3695-supervisor-grace-vt branch from 56386bc to 77e0f52 Compare July 8, 2026 07:01

muhammad-ali-e merged commit 5a385a4 into feat/UN-3445-pg-queue-integration Jul 8, 2026
4 checks passed

muhammad-ali-e deleted the fix/UN-3695-supervisor-grace-vt branch July 8, 2026 07:03

muhammad-ali-e mentioned this pull request Jul 8, 2026

UN-3695 [FIX] PG consumer — renewable lease (dead worker redelivers in minutes, not the full VT) #2154

Merged



		def shutdown_grace_from_env() -> float:
		"""Per-child graceful-drain budget (seconds) on shutdown, before SIGKILL.

Uh oh!

Conversation

muhammad-ali-e commented Jul 8, 2026

What

Why

Scope / safety

How tested

Uh oh!

coderabbitai Bot commented Jul 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review skipped

Uh oh!

muhammad-ali-e left a comment

Choose a reason for hiding this comment

Uh oh!

muhammad-ali-e Jul 8, 2026

Choose a reason for hiding this comment

Uh oh!

muhammad-ali-e Jul 8, 2026

Choose a reason for hiding this comment

Uh oh!

muhammad-ali-e Jul 8, 2026

Choose a reason for hiding this comment

Uh oh!

muhammad-ali-e Jul 8, 2026

Choose a reason for hiding this comment

Uh oh!

muhammad-ali-e Jul 8, 2026

Choose a reason for hiding this comment

Uh oh!

muhammad-ali-e Jul 8, 2026

Choose a reason for hiding this comment

Uh oh!

muhammad-ali-e Jul 8, 2026

Choose a reason for hiding this comment

Uh oh!

muhammad-ali-e commented Jul 8, 2026

Uh oh!

greptile-apps Bot commented Jul 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 5/5

Important Files Changed

Sequence Diagram

Uh oh!

muhammad-ali-e commented Jul 8, 2026

Uh oh!

Uh oh!

sonarqubecloud Bot commented Jul 8, 2026

Quality Gate passed

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

coderabbitai Bot commented Jul 8, 2026 •

edited

Loading

greptile-apps Bot commented Jul 8, 2026 •

edited

Loading