apiserver: apply the controller's jobPidInfo start-report push

handleNotifyData had no case for NotifyData_data_jobPidInfo, so the
controller's start-report push (apiserver_job_start_fun, fired from
HOOK_DETAIL_JOB_START the moment WAL_JOB_START commits) was silently
dropped. interactive_job_url then stayed invisible until the next
periodic rusage-driven job UPDATE re-delivered the full row — up to
10s later (job_rusage_collect_interval), the dominant chunk of an
-I --tty job's submit-to-first-output latency (measured ~6s on t1,
server-side work itself is ~0.6s).

Add OnJobPidInfo to the notification interface and DefaultHandler:
it merges the worker-reported pids/pgids and the exdata {"url": ...}
onto the already-stored job record (get-mutate-put, mirroring
OnJobFinished) and publishes a job.updated event. The CGO decode
side is factored into cPidInfoToRaw/pidInfoJSONB, shared with the
existing xce_job_t conversion.

Cluster-verified on t1: URL is pollable ~0.6s after submit
(previously ~11s), full submit/run/finish lifecycle clean, all
daemons healthy.

Co-Authored-By: Claude Code <noreply@anthropic.com>
This commit is contained in:
lei.chen 2026-10-15 10:08:23 +00:00
parent 66a1d1e680
commit 3545aa0d77
6 changed files with 168 additions and 16 deletions

View file

@ -115,6 +115,44 @@ func (h *DefaultHandler) OnJobFinished(jobid uint64, exitCode int) {
zap.Int("job_status", existing.JobStatus))
}
// OnJobPidInfo applies a controller-pushed job start report
// (NotifyData_data_jobPidInfo, fired from HOOK_DETAIL_JOB_START after
// WAL_JOB_START commits) onto the already-stored job record: the worker's
// pids/pgids become PIDInfo, and an interactive job's exdata {"url": ...}
// becomes InteractiveJobURL — the field xcecli's waitForInteractiveURL
// polls. Without this handler the push was silently dropped by
// handleNotifyData and the URL stayed invisible until the next periodic
// rusage-driven job UPDATE re-delivered the full row (up to 10s later,
// job_rusage_collect_interval).
//
// Mirrors OnJobFinished's get-mutate-put-publish shape: the full job row
// (submit spec, state, timestamps) always arrives ahead of or with the
// start report — SUBMIT via OnJobCreated/join push, RUN state itself via
// the dispatch-time update — so a jobid not in the store is logged and
// dropped rather than inserted half-empty.
func (h *DefaultHandler) OnJobPidInfo(jobid uint64, pidInfo *rawPidInfo) {
existing, ok := store.Jobs.Get(jobid)
if !ok {
logger.Info("OnJobPidInfo: jobid not found in store, ignoring",
zap.Uint64("jobid", jobid))
return
}
if pidInfo != nil {
if jb, ok := pidInfoJSONB(pidInfo.pids, pidInfo.pgids); ok {
existing.PIDInfo = jb
}
existing.InteractiveJobURL = interactiveURLFromExdata(pidInfo.exdata)
}
store.Jobs.Put(jobid, existing)
DefaultBroker.Publish(Event{Type: EventJobUpdated, JobID: jobid,
Status: models.JobStateStr(existing.JobStatus), Job: &existing})
logger.Info("OnJobPidInfo: start report applied",
zap.Uint64("jobid", jobid),
zap.Int("pids", len(pidInfo.pids)),
zap.String("interactive_job_url", existing.InteractiveJobURL))
}
func (h *DefaultHandler) OnQueueUpdated(queue *models.QueueInfo) {
store.Queues.Put(queue.QueueName, *queue)
}