Dafne: Documentation import

2026-08-06T13:36:46Z

Documentation import

New page

[[File:monitor_failed_task.png|thumb|right|A task in Error status]]
This page describes what to do when a task ends in '''Error''', '''Cannot connect''', '''Invalid''' or '''Warning'''.

== Finding the failure ==
Failed tasks are visible in three ways:
* the '''Error''' counter on the processing in the Active processings panel
* the status of the individual task in the task list
* the status filter chips, which restrict the list to one status group

If you know part of the error text but not which task produced it, use the '''Messages''' button in the filter bar to search the run messages directly.

== Understanding what happened ==
Select the failed task and work through its detail panel:

# '''Runs tab.''' Each execution attempt is listed with its messages. Read the message from the most recent run — this is the error returned by the target system. The badge on the tab shows how many attempts have been made.
# '''Info tab.''' Check '''Try count''' (shown as ''n of m''). If the task is configured to retry, a single failure may not be final. Compare the '''Last duration''' with the '''Avg duration''': a task that failed after running far longer than usual points at a timeout or a lock rather than a data problem.
# '''Query tab.''' The '''executed query''' shows the statement after placeholder substitution — the text actually sent to the target. This is where an incorrect parameter value becomes visible. Use '''Copy with parameters''' to reproduce the call by hand.

== Choosing the response ==

=== Retry the task ===
Use '''Repeat''' when the cause is transient and is likely to have cleared: a deadlock, a temporary shortage of resources, a target system that was briefly unavailable. The task waits for its dependencies and for resource limits exactly as it would normally.

Use '''Force repeat''' when the task should run as soon as possible without waiting for the normal constraints. Because it bypasses those constraints, it is the right choice when you have established that the blocking condition no longer applies — and the wrong choice as a routine first response.

If a task fails repeatedly for the same predictable reason, such as source data that arrives late, configure retries on the task definition rather than repeating it by hand each time. See [[Task]].

=== Mark the task as complete ===
Use '''Set complete''' when the work has been carried out another way — typically a manual correction or a load performed directly against the target. The task is recorded as ''Manual finished'' and treated as successful, so tasks depending on it will start.

This action asserts that the task's output exists. Do not use it to clear a red counter; dependent tasks will run against data that was never produced.

=== Hold the task ===
Use '''Hold''' when a task must not run for now but you do not want to disable its definition — for example while waiting for a fix, or when a source system is known to be down for the rest of the day. The task stays in the processing but is not picked up. Release it later with '''Unhold'''.

Holding is preferable to repeatedly cancelling a task that keeps being retried.

=== Cancel a running task ===
Use '''Cancel''' to stop a task that is still running — one that has hung, or that you know is working on bad input.

Whether the task actually stops depends on the connection type and on how the target system responds. Depending on that, the task may stop, may end in error, or may continue to completion if the connection type does not support cancellation. Check the status afterwards rather than assuming the cancel took effect.

=== Repair ===
'''Repair''' starts a repair run and is offered only for connection providers that support it. Where it is available, it is the appropriate response to a partially applied change, because it lets the provider resolve the inconsistency rather than re-running the whole task.

== Warnings ==
A task in '''Warning''' status has completed. The execution reported a condition configured to be treated as a warning rather than an error — for a shell script, this is a specific configured return value; see [[shell-script-connection]].

Dependent tasks are not blocked by a warning. Read the message in the Runs tab and decide whether the run needs intervention. Warnings that appear every day usually mean the threshold is set wrongly, not that there is a daily problem.

== Invalid status ==
'''Invalid''' is a technical status that occurs when the executor stops while a task is running or prepared. It does not indicate a problem with the task itself: the executor could not record how the run ended.

Before repeating an Invalid task, establish whether the work actually completed on the target system. The executor lost track of the run; the target may well have finished it. Repeating without checking can duplicate the work.

== Notifications ==
Operators assigned to a task, or to its group, are notified when it fails. Notification is sent only after the last automatic retry has been exhausted, so a task configured with retries does not generate a notification on its first failure. See [[Operator]].

== See also ==
* [[Task and processing statuses]]
* [[Investigating a stalled processing]]
* [[Reprocessing data]]

Handling failed tasks - Revision history

Dafne: Documentation import