Frozen dataclass for the number of OS threads in the process. Useful for catching leaked threads (growing unboundedly) and collapsed worker pools (a thread-pool dying under exceptions).

  • Ceiling — max_threads is a hard thread-count cap. Severity error.
  • Spike — spike_ratio + min_spike_delta detect thread leaks. Severity warning.
  • Drop — drop_ratio + min_drop_delta catch pool collapse. Severity warning.

Drop mode ships enabled by default — a pool halving is usually the signal.

Fields

NameTypeDescriptionDefault
durationstr | intShort evaluation window."1m"
baseline_durationstr | intBaseline window."15m"
max_threadsint | NoneAbsolute thread ceiling. None disables ceiling mode.100
spike_ratiofloat | NoneGrowth factor vs. baseline.1.5
min_spike_deltaint | NoneThread delta floor for a spike.10
drop_ratiofloat | NoneDecline factor in (0, 1).0.5
min_drop_deltaint | NoneThread delta floor for a drop.5

Computed properties: duration_sec, baseline_duration_sec.

Example

import snitchbot
from snitchbot import AnomalyConfig, ThreadAnomalyConfig

snitchbot.init(
    "celery-worker",
    anomaly=AnomalyConfig(
        threads=ThreadAnomalyConfig(
            max_threads=200,
            spike_ratio=2.0,
            min_spike_delta=20,
            drop_ratio=0.5,
            min_drop_delta=10,
        ),
    ),
)

Telegram shows:

🟠 anomaly · celery-worker · 44c2fb
Worker collapse: 48 -> 22 (-26)
Details
  time      11:28:11 UTC
  pid       1550
  type      threads_drop
  window    1m
  baseline  48
  current   22

Notes

InvalidAnomalyConfigError is raised on max_threads <= 0, a bad spike_ratio, or a baseline_duration shorter than duration.