QueuedEvent¶

A QueuedEvent CR is an entry in the operator's admission queue. Producers (webhook handlers, cron reconcilers) create QueuedEvent objects; the dispatcher admits them to become Task CRs when a concurrency slot is available.

apiVersion: tatara.dev/v1alpha1
kind: QueuedEvent

Internal resource

QueuedEvent is primarily an internal operator resource. You rarely create or modify them directly. They appear in kubectl get queuedevents and are useful for diagnosing why a task has not started yet.

Spec¶

Field	Type	Description
`seq`	int64	Monotonically increasing sequence number (admission order)
`class`	enum	`normal` or `alert` (alert events get reserved slots)
`kind`	string	Task kind this event will become
`autonomous`	bool	True for cron-generated (autonomous) events
`projectRef`	string	Parent Project CR name
`repositoryRef`	string	Repository CR name (empty for project-scoped kinds)
`dedupKey`	string	Dedup key; a second event with the same key is dropped if the first is still Queued or Admitted
`payload`	QueuedEventPayload	Task blueprint rebuilt verbatim on admission

QueuedEventPayload¶

Field	Type	Description
`goal`	string	Task goal text
`kind`	string	Task kind
`repositoryRef`	string	Repository CR name
`source`	TaskSource	SCM work-item that originated the event
`labels`	`map[string]string`	Labels to apply to the created Task
`annotations`	`map[string]string`	Annotations to apply to the created Task
`name`	string	Fixed Task name (for idempotent admission, e.g. issueLifecycle)
`generateName`	string	Name prefix when `name` is empty
`provider`	string	SCM provider (for pod naming)
`podRepo`	string	Repo slug used in pod name
`systemicGroup`	SystemicGroup	Systemic improvement group
`alertRule`	string	Alert rule name (incident events)

Status¶

Field	Type	Description
`state`	enum	`Queued` or `Admitted`
`taskRef`	string	Name of the Task created on admission
`admittedAt`	timestamp	When this event was admitted

Queue classes and capacity¶

normal class: up to Project.queue.capacity slots (default 3)
alert class:  up to Project.queue.alertCapacity slots (default 1)

Alert-class events (incidents) have dedicated reserved capacity. A full normal queue does not block an incoming incident from being admitted.

Dedup behavior¶

When a new QueuedEvent arrives with a dedupKey that matches an existing Queued or Admitted event, the new event is dropped. This prevents duplicate Tasks when, for example, a webhook fires twice or a cron overlaps with a still-running task for the same issue.

For issueLifecycle tasks, the dedup key is the issue owner/repo#N reference. The operator uses a fixed name (not generateName) to make admission idempotent even if the QueuedEvent is re-created.

Inspecting the queue¶

kubectl -n tatara get queuedevents
# NAME                        SEQ   CLASS    KIND             STATE
# my-project-12345            17    normal   issueLifecycle   Queued
# my-project-alert-98765      18    alert    incident         Admitted

kubectl -n tatara describe queuedevent my-project-12345

A Queued event that stays Queued for a long time indicates the queue is at capacity. Check currently running tasks:

kubectl -n tatara get tasks -l tatara.dev/project=my-project \
  --field-selector status.phase=Running