graham/prop
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Add 90-min timeout kill to PropagationGridWorker

Browse source
This commit is contained in:
Graham McIntire 2026-04-13 09:46:57 -05:00
parent 6274d03543
commit ece721e490
No known key found for this signature in database
GPG key ID: F4ABF488E6029E59
1 changed files with 13 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

14
lib/microwaveprop/workers/propagation_grid_worker.ex
View file

@ -5,10 +5,12 @@ defmodule Microwaveprop.Workers.PropagationGridWorker do
Fetches f00-f18 to provide 18-hour forecast timeline.
"""
# `unique.period` matches the 3-hour cron so a retrying job can't be
# duplicated by the next cron firing.
use Oban.Worker,
queue: :propagation,
max_attempts: 3,
unique: [period: 3600, states: [:available, :scheduled, :executing, :retryable]]
unique: [period: 10_800, states: [:available, :scheduled, :executing, :retryable]]
alias Microwaveprop.Propagation
alias Microwaveprop.Propagation.BandConfig
@ -22,6 +24,16 @@ defmodule Microwaveprop.Workers.PropagationGridWorker do
require Logger
# Hard ceiling per run. A healthy run is ~60 min (19 forecast hours × ~3 min
# each). 90 min gives 1.5× headroom and is under the 3-hour cron cadence, so
# a stuck wgrib2/HTTP call can't block the queue indefinitely. Oban kills
# the executing process on timeout, which closes linked ports and cascades
# SIGKILL to any child wgrib2 subprocess.
@run_timeout_ms 90 * 60 * 1000
@impl Oban.Worker
def timeout(_job), do: @run_timeout_ms
@max_forecast_hour 18
@impl Oban.Worker