Harden poller from Codex review: close residual storm/robustness gaps

Adversarial review (gpt-5.5) of the new architecture surfaced real gaps:

Hardware-safety / storm prevention:
- Restart-storm guard: on startup, defer the first sweep if a sweep ran
  < POLL_SWEEP_INTERVAL ago (persisted in Redis), so deploy-cycling can no
  longer trigger back-to-back full hardware sweeps (poller.py).
- Centralize pacing in the hardware gate: POLL_DRIVE_GAP is now held after
  EVERY hardware op (SMART, SES, host, MegaRAID, ledctl), not just the
  enclosure-drive loop (hwgate.py); removed the now-redundant per-loop sleeps.
- Single-instance lock made atomic (Lua compare-and-set / compare-and-expire)
  and the refresher is now fatal on any error + has a done-callback, so a
  dropped lock can never leave two pollers sweeping concurrently (store.py,
  poller.py).
- Warn loudly when POLL_CONCURRENCY > 1.

Correctness:
- Heartbeat now actually writes: cache_set omits EX when ttl<=0 (Redis
  rejects EX 0), so poller meta/last_sweep_ts persists — this also enables
  the restart-storm guard and the poller_fresh health flag (cache.py).
- Web image runs a single uvicorn worker so the MQTT publisher is a true
  singleton (two workers shared a client_id and flapped the broker) (Dockerfile).
- LED enqueue catches Redis errors -> API returns 503, not 500 (store.py).

Deploy independence:
- build.sh takes a target (web|poller|all) so web iterations never rebuild
  or repush the poller image.

Nits: drop dead SMART_CACHE_TTL constant.
This commit is contained in:
2026-06-16 15:58:43 +00:00
parent cccc2bc004
commit 7878e17e55
7 changed files with 123 additions and 44 deletions

View File

@@ -82,11 +82,15 @@ async def get_meta() -> dict | None:
# ── locate-LED command queue ────────────────────────────────────────────
async def enqueue_led(device: str, state: str) -> bool:
"""API side: push a LED request for the poller to execute. Returns False
if Redis is unavailable (so the API can surface 503)."""
if Redis is unavailable or the push fails (so the API can surface 503)."""
client = get_client()
if client is None:
return False
await client.rpush(K_LED_QUEUE, json.dumps({"device": device, "state": state}))
try:
await client.rpush(K_LED_QUEUE, json.dumps({"device": device, "state": state}))
except Exception as e:
logger.warning("LED enqueue failed: %s", e)
return False
return True
@@ -105,29 +109,37 @@ async def pop_led(timeout: int = 5) -> dict | None:
return None
# ── poller single-instance lock ─────────────────────────────────────────
# ── poller single-instance lock (atomic compare-and-set) ────────────────
# Acquire if free OR already ours, refreshing the TTL — all in one round trip
# so there's no GET/SET race that could let two pollers run concurrently.
_ACQUIRE_LUA = (
"local v = redis.call('get', KEYS[1]) "
"if v == false or v == ARGV[1] then "
" redis.call('set', KEYS[1], ARGV[1], 'EX', ARGV[2]) return 1 "
"else return 0 end"
)
# Extend only if we still own it (atomic compare-and-expire).
_REFRESH_LUA = (
"if redis.call('get', KEYS[1]) == ARGV[1] then "
" return redis.call('expire', KEYS[1], ARGV[2]) "
"else return 0 end"
)
async def acquire_lock(token: str, ttl: int = 30) -> bool:
"""Try to claim the poller lock. Returns True if acquired/owned."""
"""Atomically claim the poller lock (or re-own it). True if held."""
client = get_client()
if client is None:
return True # no Redis → no coordination possible; run anyway
got = await client.set(K_LOCK, token, nx=True, ex=ttl)
if got:
return True
current = await client.get(K_LOCK)
return current == token
return bool(await client.eval(_ACQUIRE_LUA, 1, K_LOCK, token, ttl))
async def refresh_lock(token: str, ttl: int = 30) -> bool:
"""Extend the lock if we still own it."""
"""Atomically extend the lock iff we still own it."""
client = get_client()
if client is None:
return True
current = await client.get(K_LOCK)
if current == token:
await client.expire(K_LOCK, ttl)
return True
return False
return bool(await client.eval(_REFRESH_LUA, 1, K_LOCK, token, ttl))
def now() -> float: