Repository navigation
fix(quota-plane): quota-gatherer failing every tick since 2026-11-04T18:45Z — auth.openCodeLocalCredentialSource is not a function #5638
Description
quota-gatherer.service failing every tick since 2026-11-04T18:45Z — import error auth.openCodeLocalCredentialSource is not a function
The 15-minute quota gatherer has produced no successful tick since
2026-11-04T18:30:15Z. Every tick from 18:45Z onward (18:45, 19:00,
19:15, plus extra manual/restart attempts at 19:02 and 19:22Z) fails
before any provider runs:
quota-gatherer: tick failed before providers ran: quota module import failed:
auth.openCodeLocalCredentialSource is not a function.
(In 'auth.openCodeLocalCredentialSource({ dbPath })',
'auth.openCodeLocalCredentialSource' is undefined)
Evidence (laptop amd-6, 2026-11-04 ~19:22Z):
quota-gatherer.servicestate:failed (Result: exit-code);
a manualsystemctl --user start quota-gatherer.serviceat 19:22Z
by the pool-quota steward failed identically (deterministic code
error, not a transient).- Heartbeat file
~/.local/state/pm-aggregator/quota-gatherer-heartbeat.json:
{"tickAt": "2026-11-04T19:15:06.014Z", "ok": false, ...}with the
same import error. quota-current.jsonis frozen atgeneratedAt 2026-11-04T18:30:15.509Z
(last good tick: 12 providers, 11 ok, 5,256 ms).- Timeline: the last good tick ran at checkout HEAD
45be0757b;
the first failed tick (18:45Z) ran after HEAD advanced to include
PR fix(aggregator): refresh stale quota on demand (#1171) #5263 (mergee25ebcdec8, "queue/6561-1171-quota-lazy-refresh").
The lazy-refresh change is the prime suspect for the renamed/removed
openCodeLocalCredentialSourceexport — verify against the diff.
Impact:
- The PM board has fallen back to dated fallback windows for every
pool (all rows state "stale"); the Turso hub leg no longer surfaces
at all — the code#5468 storage escalation (87.866% of the 5 GB cap
at the last good read, ~0.61 GB headroom) is invisible on the PM
pages and its attention item is gone. - The pool-quota steward is proceeding on the 18:30Z read plus live
OpenChamber/api/quota/*probes (per the standing rule, a stale
read never idles a pool), but the hub picture — including the Turso
leg, which has no live-probe route — cannot refresh until this is
fixed.
Acceptance: quota-gatherer.timer ticks succeed again
(quota-current.json generatedAt advances, heartbeat ok: true),
and the board returns to pass-through values from the fresh read.
Found by: pool-quota-steward run 155 (2026-11-04T19:2xZ), step 6(1)
read-currency check. Related: #2490 (one quota read path), #5468
(Turso escalation currently unmeasurable), #3347 (Meta read — the
live-probe route still returns null windows independently of this).
Acceptance criteria
quota-gatherer.servicecompletes a tick successfully: theauth.openCodeLocalCredentialSource is not a functionimport error is gone.quota-current.jsonis regenerated by a scheduled tick (not only a manual run) and the heartbeat reportsok: true.- The regression source is identified (prime suspect: PR fix(aggregator): refresh stale quota on demand (#1171) #5263's lazy-refresh rename/removal of the
openCodeLocalCredentialSourceexport) and the fix keeps the export surface the gatherer imports consistent with its consumers.
Activity
developing-today commentedon Nov 4, 2026
CODE-1400
dezren39 commentedon Nov 4, 2026
Steward 156th run (2026-11-04T20:21Z): quota-gatherer still failing every tick — the 20:00Z and 20:15Z timer ticks and a manual
systemctl --user start quota-gatherer.serviceat 20:21:06Z all fail identically:quota module import failed: auth.openCodeLocalCredentialSource is not a function. quota-current.json remains frozen at generatedAt 2026-11-04T18:30:15.509Z (~111 min stale at check); heartbeat ok:false. The Turso SaaS leg remains unmeasurable while the gatherer is down. The fleet proceeds on live OpenChamber /api/quota/* probes per the stale-read rule — no pool idled.dezren39 commentedon Nov 4, 2026
Intake review
auth.openCodeLocalCredentialSource is not a function);quota-current.jsonis frozen at the 18:30:15Z tick. Filed with no labels; this pass normalizes them (kind/bug, priority/high, effort/s).— fleet monitor intake, 2026-11-04
dezren39 commentedon Nov 4, 2026
Pool-quota steward, 157th run (2026-11-04T21:26Z): STILL DOWN. The 21:00Z and 21:15Z timer ticks failed identically (
auth.openCodeLocalCredentialSource is not a function), and a manualsystemctl --user start quota-gatherer.serviceat 21:25:31Z failed the same way (heartbeat ok:false, durationMs 11). quota-current.json remains frozen at generatedAt 2026-11-04T18:30:15Z (~2h55m stale). Consequence this run: the hub picture incl. the Turso leg is unmeasurable; the steward live-probed every standard pool via laptop OpenChamber :3000 instead. New unverifiable-via-hub event: opencode-go popesdrew 5h hit 100% live (was 80.96% at 20:19Z) ahead of its 5:59 PM CST reset — the hub cannot confirm fleet-visible state until this is fixed. fleet-feed-collector, pm-aggregator and openchamber all healthy.dezren39 commentedon Nov 5, 2026
Pool-quota steward run 160 (2026-11-05T00:18Z) — post-tunnel-recovery confirmation: quota-gatherer.timer is ACTIVE and ticking (~12-min cadence; last tick 18:18:02 CST), but quota-gatherer.service still fails every tick in ~20ms with the same import error: 'quota module import failed: auth.openCodeLocalCredentialSource is not a function'. quota-current.json remains frozen at generatedAt 2026-11-04T18:30:15.509Z (~5h50m stale at check time); heartbeat file is fresh but ok:false with perProvider empty. fleet-feed-collector.timer active and fleet-feed.json fresh. A restart cannot fix a deterministic import failure — this needs the code fix; the steward proceeded on live OpenChamber /api/quota probes per doctrine (a stale read never idles a pool).
dezren39 commentedon Nov 5, 2026
Steward note, 2026-11-05T01:20Z (161st run): the systemd unit still fails every tick at the same import error (
auth.openCodeLocalCredentialSource is not a function; journal ticks 18:18–19:19 CST all failed; standalone heartbeat tickAt 2026-11-05T01:19:33Z ok:false, perProvider empty).New diagnostic fact: quota-current.json nevertheless ADVANCED — generatedAt 2026-11-05T00:27:47.702Z, embedded heartbeat ok:true with per-provider rows (incl. turso via turso-api) — ~6h newer than the previously frozen 2026-11-04T18:30:15Z read. The successful 00:27:47Z write did not come from a timer tick (the 18:30:13 CST tick failed; file mtime is 18:27 CST), so a second writer/refresh path landed one good gather while the unit stays broken. The read is again aging (52 min old at this run's fetch; >20-min freshness gate), so the unit failure still stands as the blocker for a current read.
dezren39 commentedon Nov 5, 2026
Steward 162nd run (2026-11-05T02:22Z): the quota-gatherer has RECOVERED. quota-gatherer.service ticks at 2026-11-05T02:15:02Z and 02:21:29Z both completed status 0/SUCCESS (providers=12 ok=10, durationMs ~6.8-8.4s); quota-current.json generatedAt advanced to 2026-11-05T02:21:29.564Z with heartbeat ok:true. The
auth.openCodeLocalCredentialSource is not a functionimport failure that failed every tick since 2026-11-04T18:45Z no longer occurs (laptop primary ~/git/code advanced to HEAD 83f78ed, clean). Remaining per-tick errors are provider-side only: claude 'Rate limited. Retrying soon.' and meta no-data (null windows, retained last-known from 2026-11-04T18:30Z; tracked on #3347). Steward did not restart anything — recovery arrived via the code path. This issue's failure mode appears resolved; closing is the fix owner's call.dezren39 commentedon Nov 5, 2026
Agent note - handed off
ses_e5e5345be21ffelgIo1QE32axBdezren39 commentedon Nov 5, 2026
Triage — queue task 10093
The operational failure is resolved. Steward run 162 records scheduled
quota-gatherer.serviceticks succeeding at 2026-11-05T02:15:02Z and 02:21:29Z,quota-current.jsonadvancing to 2026-11-05T02:21:29.564Z, and heartbeatok:true(recovery evidence).Current
mainisf0fa32003fe5399ff6bc117d08ecc80a7322250d. The durable code path in PR #6136 (change97f8871b8b201f658cafba00bf2fd327272e862a, merged as97cd55c) replaces the gatherer's in-process OpenChamber auth/quota import withopencode quota --json; the missingopenCodeLocalCredentialSourceexport is no longer called. The suspected PR #5263 did not touch the gatherer/auth import and merged after the reported first failure. The runtime recovery was already reported at checkout83f78ed2347e, before PR #6136; repository history does not explain that immediate recovery, so its cause remains unconfirmed.The remaining Claude rate-limit and Meta no-data conditions are provider-side and tracked separately on #3347. Since scheduled ticks recovered and current main removes the failing dependency, closing #5638 as completed; no blocker remains for this failure mode.