How can we adjust TTL idle_seconds for Janitor close_cold (Docker Job Queue API)?

[pbtsrc]

In case of large complex jobs we often get errors like in #1842 and it is looks like the idle_seconds is too low for or tasks. How can we increase the TTL/idle_seconds?

Example:

🧹 Janitor Events
11:15:49 🧹❄️ close_cold sig=21af23fd {"idle_seconds":311,"ttl":300}

[hafezparast]

The janitor's TTL is configurable via config.yml. In your Docker deployment, edit the config:

crawler:
  pool:
    idle_ttl_sec: 900   # increase from default 300s (5 min) to 15 minutes

How the janitor works:
The TTLs are adaptive based on container memory pressure:

Memory usage	Cold pool TTL	Hot pool TTL	Check interval
> 80%	30s	120s	every 10s
> 60%	60s	300s	every 30s
< 60%	idle_ttl_sec (default 300s)	idle_ttl_sec * 2	every 60s

Your log shows idle_seconds: 311, ttl: 300 — meaning memory is below 60% and the cold TTL is using the default of 300s. Increasing this value will give your long-running jobs more breathing room.

Note: Under high memory pressure (>80%), the janitor aggressively closes idle browsers regardless of your idle_ttl_sec setting (cold TTL drops to 30s). If you're hitting that too, consider increasing the container's memory allocation.

(Side note: the comment in the default config.yml says "30 min janitor cutoff" but the actual value 300 is 5 minutes — that's a typo in the config comment.)