Pipeline dashboard#249
Conversation
dda02e0 to
78e81e5
Compare
|
Hmmmmm, most of these stats should be accessible via querying the database via Can we start there, then consider telemetry later? |
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Run black formatter on deploy.py, views.py, migration - Add nosec B608 to f-string SQL with hardcoded table names - Fix migration chain: down_revision -> e5f9a2c7d834 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Narrow _REGISTRY type with an early return guard instead of assert (which bandit flags as B101). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
e85d8dd to
699879e
Compare
|
splitting push gateway and metrics stuff |
The pipeline dashboard derives all stats from existing DB columns (created, updated, built, error, complete). The prometheus_client pushgateway integration can be added later when needed. Removes: cabotage/celery/metrics.py, all record_*_metrics() calls, PUSHGATEWAY_URL config. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
|
Pushgateway telemetry split out to #319 for later pickup. |
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Spreads images/releases/deployments across 72h with varied build durations so the pipeline dashboard shows meaningful charts and stats. Also fixes seed login message (port 8000, correct credentials). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
SQLAlchemy's Timestamp mixin overwrites `updated` on commit. Defer timestamp assignments to raw UPDATE statements after the final commit so pipeline dashboard shows realistic durations. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Spreads deploys across 90 days so 7d/30d/90d range filters show different data. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
c795fe6 to
a9f6d8f
Compare
Reverts TESTING guards in render_process_container and render_podspec, and removes REGISTRY_PULL_K8S/TESTING config that were local dev workarounds, not pipeline features. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
|
@ewdurbin i think this is a lil bit smaller and nicer now, only using the db stuff we have. doing a final passthrough check on the py code but could use your review for a merge |
The deferred _timestamp_fixups machinery was unnecessary — SQLAlchemy's `default=` only fires when no value is provided, so setting created/updated directly after flush() works fine. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Code Review — Pipeline DashboardCross-check review via Codex (gpt-5.4, xhigh effort). Must Fix
Should Fix
Confirmed OK
|
i think itd be better to move scripts up a level outside of the real cabotage module code in another PR |
Removes the broken defer/fixup mechanism. Instead, backdates all pipeline timestamps via raw SQL after the final ORM commit, bypassing SQLAlchemy's onupdate handlers. 15 runs across 90 days. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Description
Enables the Pipeline tab
Closes #247