hermes-agent/tests/hermes_cli/test_gateway_runtime_health.py
Cossackx ad7b4c7d2e test(gateway): cover stale gateway_state.json detection (TTL + PID liveness)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-21 12:39:59 -07:00

166 lines
5.7 KiB
Python

from datetime import datetime, timedelta, timezone
from hermes_cli.gateway import _runtime_health_lines
def _iso_age(seconds_ago: float) -> str:
"""ISO-8601 UTC timestamp ``seconds_ago`` in the past (drives _marker_is_stale)."""
return (datetime.now(timezone.utc) - timedelta(seconds=seconds_ago)).isoformat()
_STALE_LINE_PREFIX = "⚠ Stale gateway_state.json:"
def _stale_lines(lines):
return [ln for ln in lines if ln.startswith(_STALE_LINE_PREFIX)]
def test_runtime_health_lines_flags_stale_running_with_dead_pid(monkeypatch):
"""Stale updated_at + dead PID + 'running' -> contradiction line, no draining line."""
from gateway import status as status_mod
monkeypatch.setattr(
"gateway.status.read_runtime_status",
lambda: {
"gateway_state": "running",
"pid": 4242,
"start_time": 111,
"updated_at": _iso_age(600), # well past the 120s TTL -> stale
"active_agents": 0,
},
)
# Recorded PID is gone (ungraceful kill); no real process is touched.
monkeypatch.setattr(status_mod, "_pid_exists", lambda pid: False)
monkeypatch.setattr(status_mod, "_get_process_start_time", lambda pid: None)
lines = _runtime_health_lines()
stale = _stale_lines(lines)
assert len(stale) == 1, lines
assert "recorded state 'running'" in stale[0]
assert "recorded process is gone" in stale[0]
# The misleading live-state summary must be suppressed.
assert not any("draining" in ln.lower() for ln in lines), lines
def test_runtime_health_lines_no_warning_for_fresh_running_with_live_pid(monkeypatch):
"""Fresh updated_at + live PID + 'running' -> no contradiction line."""
from gateway import status as status_mod
monkeypatch.setattr(
"gateway.status.read_runtime_status",
lambda: {
"gateway_state": "running",
"pid": 4242,
"start_time": 111,
"updated_at": _iso_age(5), # fresh, within the 120s TTL
"active_agents": 2,
},
)
monkeypatch.setattr(status_mod, "_pid_exists", lambda pid: True)
# start_time matches the record -> not a recycled PID.
monkeypatch.setattr(status_mod, "_get_process_start_time", lambda pid: 111)
lines = _runtime_health_lines()
assert _stale_lines(lines) == [], lines
def test_runtime_health_lines_no_warning_for_stale_running_but_live_pid(monkeypatch):
"""Stale updated_at but PID still live (long-idle gateway) -> no false positive."""
from gateway import status as status_mod
monkeypatch.setattr(
"gateway.status.read_runtime_status",
lambda: {
"gateway_state": "running",
"pid": 4242,
"start_time": 111,
"updated_at": _iso_age(600), # stale timestamp...
"active_agents": 0,
},
)
# ...but the recorded process is genuinely alive (start_time matches).
monkeypatch.setattr(status_mod, "_pid_exists", lambda pid: pid == 4242)
monkeypatch.setattr(status_mod, "_get_process_start_time", lambda pid: 111)
lines = _runtime_health_lines()
# Guard requires BOTH stale AND dead PID; live PID must suppress the warning.
assert _stale_lines(lines) == [], lines
def test_runtime_health_lines_treats_missing_updated_at_as_stale(monkeypatch):
"""Missing updated_at (degrade path) + dead PID + 'running' -> contradiction line."""
from gateway import status as status_mod
monkeypatch.setattr(
"gateway.status.read_runtime_status",
lambda: {
"gateway_state": "running",
"pid": 4242,
"start_time": 111,
# no "updated_at" -> runtime_status_is_stale must treat as stale
"active_agents": 0,
},
)
monkeypatch.setattr(status_mod, "_pid_exists", lambda pid: False)
monkeypatch.setattr(status_mod, "_get_process_start_time", lambda pid: None)
lines = _runtime_health_lines()
stale = _stale_lines(lines)
assert len(stale) == 1, lines
assert "recorded state 'running'" in stale[0]
def test_runtime_health_lines_include_fatal_platform_and_startup_reason(monkeypatch):
monkeypatch.setattr(
"gateway.status.read_runtime_status",
lambda: {
"gateway_state": "startup_failed",
"exit_reason": "telegram conflict",
"platforms": {
"telegram": {
"state": "fatal",
"error_message": "another poller is active",
}
},
},
)
lines = _runtime_health_lines()
assert "⚠ telegram: another poller is active" in lines
assert "⚠ Last startup issue: telegram conflict" in lines
def test_runtime_status_running_pid_validates_live_gateway_record(monkeypatch):
from gateway import status as status_mod
runtime = {
"pid": 12345,
"kind": "hermes-gateway",
"argv": ["/opt/hermes/hermes_cli/main.py", "gateway", "run", "--replace"],
"start_time": None,
"gateway_state": "running",
}
monkeypatch.setattr(status_mod, "_pid_exists", lambda pid: pid == 12345)
monkeypatch.setattr(status_mod, "_get_process_start_time", lambda pid: None)
monkeypatch.setattr(status_mod, "_looks_like_gateway_process", lambda pid: False)
assert status_mod.get_runtime_status_running_pid(runtime) == 12345
def test_runtime_status_running_pid_rejects_stopped_record(monkeypatch):
from gateway import status as status_mod
runtime = {
"pid": 12345,
"kind": "hermes-gateway",
"argv": ["/opt/hermes/hermes_cli/main.py", "gateway", "run", "--replace"],
"gateway_state": "stopped",
}
monkeypatch.setattr(status_mod, "_pid_exists", lambda pid: True)
assert status_mod.get_runtime_status_running_pid(runtime) is None