Skip to main content

Background Services

Dashboard view of WarpBackgroundService instances — long-running in-process work managed by Warp (restart-on-fault, optional cluster-singleton coordination, captured ILogger<T> output). The Services nav entry is always available; the list page is empty when no service is registered.

List Page​

One row per registered service name, aggregated across all servers running that service.

  • Name — the WarpBackgroundService.Name (defaults to the CLR type name).
  • Scope badge — Per Server or Singleton.
  • Status summary — for per-server: Running 3/3; for singleton: Running on warp-demo-server, 2 waiting.
  • Restart count — sum across all instances of this service.
  • Last error type — the exception type from the most recent fault if any instance is currently Faulted. Empty otherwise.
  • Configuration mismatch badge — appears when a rolling deploy ships a service with a different Scope than the existing Definition row. Warp refuses to run user code on the mismatched server until the deploy converges.

Polls every ~2 s while open. Services are ordered by FirstSeenAt (oldest registered first) so the list stays stable across refreshes.

Detail Page​

Per-service drill-down. The header shows the scope, when the service was first registered, and when it was last seen.

Per-Instance Tabs​

One tab per server with an Instance row for this service. Tab labels show the server name (resolved through the EF nav property; the Guid is still surfaced as a secondary line for debugging). Tabs are ordered by StartedAt — longest-running instance first — so the order stays stable across polls.

Each tab shows:

  • Status — Running, Waiting, Faulted, Restarting, or ConfigurationMismatch.
  • Started at — when this supervisor entered the current Running window.
  • Last heartbeat — refreshed every ~3 s by the Heartbeat server task.
  • Restart count — current accumulated count (resets to 0 after a successful run of ≥ 5 minutes).
  • Last error — full captured exception when the instance is in Faulted state. Already 4 KB-capped at capture time, so no further truncation in the dashboard.

Lease Panel (Singleton scope only)​

For services with Scope = Singleton, a Lease panel renders below the instance tabs:

  • Holder — the server name currently holding the lease.
  • Holder server ID — the Guid for debugging / log correlation.
  • Expires — live TTL countdown to LeaseExpiresAt. The Heartbeat task on the holder renews this every ~3 s; if the holder dies, the lease expires after ≤ 30 s and another server acquires.

Per-server services don't render this panel (each instance is independent).

Log Tail​

Combined log view across all instances of this service.

  • Source filter — All, Lifecycle (events emitted by the supervisor: Started, LeaseAcquired, LeaseLost, Faulted, Restarting, Stopped, ConfigurationMismatch), or User (entries from your ILogger<T> calls inside ExecuteAsync).
  • Level filter — minimum severity from Information+ up to Critical.
  • Server column — readable server name per row (Guid available on hover).
  • Expandable exceptions — click any entry with ExceptionType set to reveal the full exception message inline.

Polls incrementally every ~2 s using the highest seen log id as a cursor, so the tail stays at the head without refetching old rows.

For per-server services the same log table is shown — entries from each server's instance are interleaved chronologically with a Server column to disambiguate.

For full feature documentation — scope semantics, lease coordination details, captive-scoped-dependency foot-gun, log capture guardrails, telemetry counters — see Background Services.