#1065 Fix for console and file explorer timeout Pull request opened by javuto Console & File Explorer Timeout Fix Problem Statement Console commands (and file explorer requests) submitted while a node was still "warming up" its accelerated polling could expire undelivered, forcing operators to retry until acceleration happened to kick in. Root Cause

warmupQueryWait

(shared by both console and file explorer) predicted the node's next poll from

node.LastQueryRead

. That column is stamped by osctrl-tls's batch writer, which coalesces check-ins and flushes on

--writer-timeout

(default 60s — the same magnitude as the default 60s poll interval). A stamp lagging one flush cycle made a live, mid-cycle node look "overdue": •

wait <= 0

→ zero extra wait granted • Warmup command expired before the node's real next read → timeout The refinement carried error ±interval, i.e. it had no reliable signal. Changes Backend (benefits console and file explorer) •

cmd/api/handlers/console.go

—

warmupQueryWait

no longer trusts the recorded read; it always reserves a full poll interval (capped by

maxWarmupWait

). Delivery is guaranteed for live nodes; once acceleration is active, commands complete on the first fast poll, so the extra expiration only bounds how long a dead node's command stays pending. •

cmd/api/handlers/file_explorer.go

— same helper, so list/stat/priming get identical protection. • Removed the now-unneeded per-submit node lookup (

sessionNode

) — one fewer DB query on every console command / file explorer submit. Frontend (console only) •

frontend/src/features/nodes/NodeConsolePage.tsx

— SSE

resource.changed

events now also invalidate the priming poll (its query key was previously missed), so the "warming" badge and header metadata resolve immediately when the node answers instead of lagging a reconcile interval. Security / Operational Impact • No auth, permission, or route changes; handlers remain thin • Longer expirations only affect pending duration of hidden internal queries; results are still resolved via existing terminal-status handling • Dead nodes still fail fast: base timeout + bounded warmup wait (

maxWarmupWait = 10m

) Validation | Check | Result | | ---------------------------- | ---------- | | go test ./... | ✅ all pass | | Frontend npm test | ✅ 416/416 | | npm run check (tsc) | ✅ clean | | golangci-lint (changed pkgs) | ✅ 0 issues | | make openapi-check | ✅ current | Regression Tests Added •

console_test.go

/

file_explorer_test.go

— stale

last_query_read

(overdue stamp) must not shrink the warmup window; expiration must cover a full interval •

NodeConsolePage.test.tsx

— a priming live-update event triggers a refetch of the priming command jmpsec/osctrl