PrefectHQ/prefect - CodSpeed

prefect

Blog Docs Changelog

Performance History

Latest Results

Fix worker drain heartbeat and submission races

qdivan:fix/process-worker-sigterm-drain-22528

36 minutes ago

Keep worker startup alive when the Prefect API is unreachable (#22640) Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> Co-authored-by: Alex Streed <alex.s@prefect.io>

main

3 hours ago

Merge remote-tracking branch 'origin/main' into worker-start-background # Conflicts: # src/prefect/cli/_worker_utils.py

SreeramaYeshwanthGowd:worker-start-background

3 hours ago

feat: show tooltips for flow run states (#22617)

main

3 hours ago

chore: retrigger CI The CLI startup benchmark flagged `prefect dev --help` as a false-positive regression unrelated to this PR's changes (worker.py/_worker_utils.py/ _server_utils.py are lazily imported and dev.py has no import path to them). Empty commit to get a fresh benchmark run.

SreeramaYeshwanthGowd:worker-start-background

3 hours ago

Verify worker identity, wait for exit, and reserve names atomically Addresses review feedback on the background worker PID tracking. Forward the active profile to the detached worker so it polls the same API the parent validated against, instead of reloading whatever profile happens to be active on disk when it starts. Verify process identity, not just liveness, before signaling a worker. Each PID file is paired with a start time marker recorded when the worker was spawned, so a PID the OS later reuses for an unrelated process is treated as stale rather than signaled. Wait for a worker to actually exit after signaling it before reporting success and dropping its PID file, since a worker treats the first signal as a graceful shutdown request and may keep running while in flight runs finish. A worker still running after the grace period is reported as still shutting down and remains tracked. Reserve a worker name atomically before spawning, so two concurrent background starts for the same name cannot both pass a check and then race to write the PID file. Normalize tracked workers before deciding what to stop: stale or corrupt PID files are cleaned up first and no longer count toward "multiple workers are running" ambiguity.

SreeramaYeshwanthGowd:worker-start-background

4 hours ago

Address worker startup recovery feedback

devin1/oss-8106-worker-startup-api-unreachable

6 hours ago

Fix worker startup recovery lifecycle

devin1/oss-8106-worker-startup-api-unreachable

9 hours ago

Latest Branches

0%

Gracefully drain active worker runs on termination signals#22580

50 minutes ago

13760a8

qdivan:fix/process-worker-sigterm-drain-22528

0%

Add prefect worker start --background and prefect worker stop#22588

3 hours ago

b59d9a6

SreeramaYeshwanthGowd:worker-start-background

0%

Keep worker startup alive when the Prefect API is unreachable#22640

6 hours ago

7b04d2b

devin1/oss-8106-worker-startup-api-unreachable

© 2026 CodSpeed Technology

Home Terms Privacy Docs