Skip to content

Results and tasks

A task is the atomic unit of execution on a telescope. Your observation is first broken into results: one per sample of each request in each epoch (and per copy, if you asked for copies). Each result is a deliverable that stays put until it is fulfilled or canceled. To get it taken, Skynet’s dispatcher creates an ObservationTask for it: “this result, on this specific telescope and instrument.” The telescope’s scheduler then gives the task a planned start time inside an observing window.

On the observation detail page, each result on the Results page lists its task (or tasks) underneath it.


Each task under a result shows:

Item What it means
Status dot The task’s status (see below)
Task number Task identifier — useful when correlating with telescope logs
Telescope Which telescope the dispatcher assigned the task to
Scheduled When the telescope’s scheduler plans to start it
Started When the telescope actually started it
Completed When it finished; until then, the status in words

A task’s status is one of:

Status Meaning
dispatched Assigned to a telescope, waiting for that telescope’s scheduler to place it
pending Placed in an observing window; the scheduler may still move it to a later window
active Running on the telescope now
completed Finished successfully
interrupted Stopped before finishing, keeping the frames it had already captured

The page refreshes about every 20 seconds while the observation is active.


Each request in your observation produces many results, and each result usually has exactly one task:

  • Epochs — every epoch produces another set of results.
  • Samples — every sample of a request is its own result. For a series request, each sample is one whole series.
  • Copies — with All eligible instruments or A fixed number of instruments, each copy is its own result, so each gets its own task.

Tiling does not create extra tasks: when an optical target needs several fields of view, the tiles are part of the one task. A radio mapping pattern (daisy or raster) is likewise one task.

A result can collect more than one task over its life. If a telescope cannot finish the work (it misses its window, goes offline for a long time, or can no longer serve the request), the dispatcher hands the result back and creates a new task, preferring the instrument the work was interrupted on. A task that already captured frames is kept as interrupted; one that captured nothing is simply replaced. Only one task per result is live at a time.


Skynet has no “failed” task status. Instead:

  • Window passes — a task that is not yet running when its observing window closes is moved to a later window by the telescope’s scheduler (or handed back by the dispatcher if the telescope keeps missing it); the result is not abandoned.
  • Telescope can’t serve it — the dispatcher hands the result back and dispatches a new task, possibly to another telescope.
  • Interrupted mid-task — the task ends as interrupted, its frames are kept, and the result’s remaining work goes to a continuation task.
  • Deadline passes — when the observation’s Cancel on date passes, results that are still outstanding are canceled. On the Results page, a result canceled before any capture reads Not taken; one with partial data reads Interrupted.

The telescope reports why a task is in its current state (for example, the sky is too bright), and the scheduler records why a task is unplaced. If a task keeps stalling the same way, that’s a telescope-side problem worth flagging to the owner.


Task IDs are useful when:

  • Correlating with telescope logs — the telescope owner can use the task ID to find the exact log lines for that task.
  • Filing a problem report — quoting the task ID gives operators the exact context to investigate.
  • Pulling data through the API — every captured frame records the task that took it (ObservationAsset.task_info), which is how files trace back to the task that produced them. The Data page can be filtered to one task’s frames.