mirror of
https://gh-proxy.org/https://github.com/klinsmaya/multica.git
synced 2026-11-08 09:26:50 +08:00
bcf40040e0
* fix(agent): join Pi cancellation cleanup and guard continuation Co-authored-by: multica-agent <[email protected]> * fix(agent): preserve Pi completion and negotiate cleanup Co-authored-by: multica-agent <[email protected]> * perf(agent): avoid Pi exit delay and index cleanup guards Co-authored-by: multica-agent <[email protected]> --------- Co-authored-by: Multica Eve <[email protected]> Co-authored-by: multica-agent <[email protected]>
73 lines
3.5 KiB
SQL
73 lines
3.5 KiB
SQL
-- name: TaskBlockedByProcessCleanup :one
|
|
-- A cleanup failure fences this agent on this runtime, including already queued
|
|
-- work and future scheduled runs. Only a NEW explicit manual rerun can cross
|
|
-- the fence; its successful completion allows later automated work again.
|
|
-- Existing failure and result fields are the durable safety markers.
|
|
SELECT EXISTS (
|
|
SELECT 1
|
|
FROM agent_task_queue candidate
|
|
JOIN LATERAL (
|
|
SELECT CASE WHEN unsafe.status = 'cancelled'
|
|
AND unsafe.result->>'process_cleanup_pending' = 'true'
|
|
THEN unsafe.completed_at + interval '30 seconds'
|
|
ELSE COALESCE(unsafe.completed_at, unsafe.created_at)
|
|
END AS completed_at
|
|
FROM agent_task_queue unsafe
|
|
WHERE unsafe.agent_id = candidate.agent_id
|
|
AND unsafe.runtime_id = candidate.runtime_id
|
|
AND (unsafe.failure_reason = 'cleanup_unconfirmed'
|
|
OR (unsafe.status = 'cancelled' AND unsafe.result->>'process_cleanup_pending' = 'true'
|
|
AND unsafe.completed_at <= now() - interval '30 seconds'))
|
|
ORDER BY unsafe.completed_at DESC, unsafe.id DESC
|
|
LIMIT 1
|
|
) unsafe ON true
|
|
WHERE candidate.id = @task_id
|
|
AND candidate.status IN ('queued', 'dispatched', 'waiting_local_directory')
|
|
AND NOT EXISTS (
|
|
SELECT 1 FROM agent_task_queue manual
|
|
WHERE manual.agent_id = candidate.agent_id
|
|
AND manual.runtime_id = candidate.runtime_id
|
|
AND manual.rerun_of_task_id IS NOT NULL
|
|
AND manual.retry_of_task_id IS NULL
|
|
AND manual.created_at > unsafe.completed_at
|
|
AND (manual.id = candidate.id OR manual.status = 'completed')
|
|
)
|
|
);
|
|
|
|
-- name: FailCancelledTaskProcessCleanup :one
|
|
-- The daemon's cancellation acknowledgement can discover that local work did
|
|
-- NOT stop. Preserve the cancelled run's metadata but surface this as failure.
|
|
-- The status CAS makes replay harmless and leaves completed/failed rows alone.
|
|
UPDATE agent_task_queue
|
|
SET status = 'failed',
|
|
error = concat_ws('; ', NULLIF(error, ''), @error_message::text),
|
|
failure_reason = 'cleanup_unconfirmed',
|
|
completed_at = now(),
|
|
chat_finalize_deferred_at = NULL
|
|
WHERE id = @id AND status = 'cancelled'
|
|
RETURNING *;
|
|
|
|
-- name: MarkPiTaskProcessCleanupPending :exec
|
|
-- Cancellation makes the task terminal before the daemon finishes stopping it.
|
|
-- Only a run that negotiated confirmation at start holds a fence until ack.
|
|
UPDATE agent_task_queue task
|
|
SET result = COALESCE(task.result, '{}'::jsonb) || '{"process_cleanup_pending":true}'::jsonb
|
|
WHERE task.id = @id AND task.status = 'cancelled' AND task.started_at IS NOT NULL
|
|
AND task.result->>'process_cleanup_capability' = 'pi-process-cleanup-v1'
|
|
AND EXISTS (SELECT 1 FROM agent_runtime runtime WHERE runtime.id = task.runtime_id AND runtime.provider IN ('pi', 'omp'));
|
|
|
|
-- name: ClearTaskProcessCleanupPending :exec
|
|
UPDATE agent_task_queue
|
|
SET result = result - 'process_cleanup_pending'
|
|
WHERE id = @id AND status IN ('cancelled', 'failed')
|
|
AND result->>'process_cleanup_pending' = 'true';
|
|
|
|
-- name: TaskWaitingForProcessCleanup :one
|
|
SELECT EXISTS (
|
|
SELECT 1 FROM agent_task_queue candidate
|
|
JOIN agent_task_queue stopping ON stopping.agent_id = candidate.agent_id AND stopping.runtime_id = candidate.runtime_id
|
|
WHERE candidate.id = @id AND candidate.status IN ('dispatched', 'waiting_local_directory')
|
|
AND stopping.status = 'cancelled' AND stopping.result->>'process_cleanup_pending' = 'true'
|
|
AND stopping.completed_at > now() - interval '30 seconds'
|
|
);
|