valka

Web Dashboard

Real-time monitoring dashboard for tasks, workers, events and the cluster.

Valka includes a built-in web dashboard that is served directly by the Valka server. No additional setup is required. The dashboard is available as soon as the server starts.

Accessing the Dashboard

Open your browser and navigate to the Valka server's HTTP address:

http://<your-server>:8989

The dashboard is a single-page application served alongside the REST API. There is no separate process or deployment step. In a cluster, every node serves it, and every node shows the whole cluster: open it on any node.

Pages

Dashboard

The main overview page showing:

  • Task counts by status (total, pending, running, completed, failed)
  • A cluster strip: health, nodes alive, shards owned, unflushed WAL records, dirty shards
  • Queue overview
  • Live event feed

Tasks

Browse and filter all tasks:

  • Filter by queue and status
  • Create a task from a form
  • Open any task's details

Task Detail

Deep dive into a specific task:

  • Overview: status, input/output, metadata, timing, and the shard that holds the task
  • Runs: all execution attempts with worker, node and duration
  • Steps: checkpoints recorded by step helpers, with the attempt that recorded each one
  • Logs: real-time structured log viewer per run
  • Signals: signal history with delivery status, and a form to send one

Workers

Monitor connected workers on every node:

  • Worker name, node, queues, and concurrency
  • Active task count
  • Connection uptime
  • Last heartbeat timestamp

Cluster

Nodes, shard ownership and write-ahead log health, in four tabs (Overview, Shards, Storage, Operations) and a page per node:

  • Overview: a health banner (ok, degraded or critical, with the reasons: unowned shards, suspect or poisoned nodes, nodes that did not answer), one card per node (status, incarnation, shards owned, pending and running tasks, workers, unflushed WAL records), and a membership table built from the leases in the bucket: each node's incarnation, lease state, lease age, eligibility for placement, protocol version, unsplit logs and operator pins. Nodes that died stay visible here, known only from their lease.
  • Node page: one node's WAL (incarnation, durable and next log position, unflushed records, oldest unacknowledged write, writer state), its snapshots (last round, dirty shards, oldest uncovered log position) with a Snapshot now button on the node serving the page, and the workers connected to it.
  • Shards: a heatmap of every shard, colored by owner, pending load, records since the last snapshot or snapshot age. Click a shard for its owner, generation, phase, task counts per queue, sequence numbers and latest snapshot. Filter by node.
  • Storage: WAL segments, snapshots and task log chunks in the bucket, an upper-bound estimate of daily PUT cost, and the bucket layout.
  • Operations: drain a node (it hands its shards over, then stops), move a shard to a node and pin it there, unpin shards, change a queue's partition count, and the audit log of every admin action. These need the admin token (VALKA_ADMIN_TOKEN), which you enter on the page; it is kept for that browser tab only. Every destructive action asks for confirmation first.

Watch a takeover here: kill a node and see it turn suspect, then dead, while its shards move to the other nodes. See Clustering.

Signing in

When the server is started with VALKA_API_KEYS, the dashboard asks for one of those keys the first time a request is refused, checks it with the server, and keeps it in your browser until you choose Sign out in the sidebar.

Events

Real-time stream of task events: every task status change, from every node.

Uses Server-Sent Events (SSE) for live updates without polling.

Dead Letters

Browse tasks that exhausted all retries:

  • Original task details and input
  • Error messages from failed attempts
  • Attempt count and timing

On this page