Xray
1h 6h 24h 48h 3d 5d 7d 14d 30d ‹
Last 1 hour▾
Ends now. Anything up to this cluster's 30-day retention.
Start and end times
Cancel
Interpreted in UTC (change it at the right of this bar). Earliest available: 2026-10-19 13:42:34.
› →|
updated 1d ago
Sample data. This database was generated by scripts/seed_dev_db.py — every service, deploy, person and figure on this page is synthetic. Names are sample.*@seed.invalid and image tags are prefixed SEED-. Nothing here happened.
← findings

workplan.summarization-sweep: 30 failed runs in the last 6 h — checkpoint scoring: unknown checkpoint source type ?

critical Background job failing master open impact 61 opened 26d ago · last observed 1d ago · observed 545× · #5483 · c8c4cddd279a421a0b6c

What happened template

The background job workplan.summarization-sweep on master failed 30 runs recently (30 failure lines; lines within one run count once), detected via its structured error/exception field (below ERROR severity). A job that logs its failures below ERROR severity — a WARNING carrying an error field — is invisible to the novel-shape detector and to the page's own failed count; this detector reads the error field itself, which is why the finding now appears. The sample failure is in the facts.

A named background job whose runs keep failing (error-severity bursts) in the last hour — volume, not novelty; the novel-shape detector only fires the first time.

First look — investigated before anyone looked

workplan.summarization-sweep: 28 failed runs in the last 6 h — checkpoint scoring: unknown checkpoint source type ?.
Cause not clear from logs and recent changes; checked: error lines (none stored), deploys in the 24 h before (12), commits in the deploy's range (0), this finding's history, other findings at the same time (1).

low confidence

claude-opus-5-5 · 1d ago · $0.04 — every link above is evidence it read; nothing else was claimed. the model named pubsub/checkpoint_scoring_sweep.go:115, which the evidence does not hold — its answer was withheld.

The facts

jobworkplan.summarization-sweep
serviceworkplansvc
failed_runs30
error_lines30
samplecheckpoint scoring: unknown checkpoint source type "clinical_trial_data"
error_signaturecheckpoint scoring: unknown checkpoint source type ?
failing_atpubsub/checkpoint_scoring_sweep.go:115
trace—
detected_viastructured error/exception field (below ERROR severity)
notea run is a burst of lines that truly look like failures — an error/exception field, an error-worded message, or a failure status; lines within one burst count as one run. Lines a job dumps at ERROR that are neither (the steps that SUCCEEDED, banner rules) are ignored, so a service that logs its whole context on failure is not miscounted. `failing_at` is the call site those failures came from — for a coarse job (one set_job wrapping a pipeline) more specific than the job name.
discriminator
window2026-11-17 04:40:41 UTC → 2026-11-17 10:40:41 UTC

The story — No workplansvc deploy or hand change is recorded in the hour before this opened. Everything below is workplansvc only.

History

  • 2026-11-16 17:27:13 UTCescalated x-ray · severity=critical was=warning title=workplan.summarization-sweep: 28 failed runs in the last 6 h — checkpoint scoring: unknown checkpoint source type ? level=high was_level=high finding={'id': 5483, 'fingerprint': 'c8c4cddd279a421a0b6c', 'kind': 'job.failing', 'severity': 'critical', 'cluster': 'master', 'subject': 'workplan.summarization-sweep', 'title': 'workplan.summarization-sweep: 28 failed runs in the last 6 h — checkpoint scoring: unknown checkpoint source type ?', 'facts': {'job': 'workplan.summarization-sweep', 'service': 'workplansvc', 'failed_runs': 28, 'error_lines': 28, 'sample': 'checkpoint scoring: unknown checkpoint source type "clinical_trial_data"', 'error_signature': 'checkpoint scoring: unknown checkpoint source type ?', 'failing_at': 'pubsub/checkpoint_scoring_sweep.go:115', 'trace': None, 'detected_via': 'structured error/exception field (below ERROR severity)', 'window': [1794828433, 1794850033], 'note': 'a run is a burst of lines that truly look like failures — an error/exception field, an error-worded message, or a failure status; lines within one burst count as one run. Lines a job dumps at ERROR that are neither (the steps that SUCCEEDED, banner rules) are ignored, so a service that logs its whole context on failure is not miscounted. `failing_at` is the call site those failures came from — for a coarse job (one set_job wrapping a pipeline) more specific than the job name.', 'discriminator': ''}, 'evidence': [{'label': "This job's page", 'href': '/jobs/workplan.summarization-sweep?cluster=master&from=1794828433&to=1794850033'}, {'label': 'Background jobs', 'href': '/jobs?cluster=master&from=1794828433&to=1794850033'}, {'label': 'Its error logs', 'href': '/logs?cluster=master&from=1794828433&to=1794850033&severity=ERROR&q=workplan.summarization-sweep'}], 'provenance': {'source': 'live', 'collected_at': None, 'age_seconds': None, 'stale': False, 'shown': None, 'total': None, 'truncated': False, 'coverage_from': None, 'retain_days': None, 'note': 'error-severity job lines from BigQuery, last hour, burst-grouped per job'}, 'impact': 56.0, 'status': 'open', 'first_seen': 1792762426, 'last_seen': 1794850033, 'resolved_at': None, 'acked_by': None, 'acked_at': None, 'snoozed_until': None, 'occurrences': 529, 'narrative': "The background job workplan.summarization-sweep on master failed 23 runs recently (23 failure lines; lines within one run count once), detected via its structured error/exception field (below ERROR severity). A job that logs its failures below ERROR severity — a WARNING carrying an error field — is invisible to the novel-shape detector and to the page's own failed count; this detector reads the error field itself, which is why the finding now appears. The sample failure is in the facts.", 'narrative_kind': 'template', 'discriminator': '', 'level': 'high', 'told_level': None, 'episode_at': 1792762426, 'chat_thread': None, 'kind_label': 'Background job failing', 'group': 'Jobs', 'age_s': 2087635, 'stale_s': 28}

Check it yourself

Every link carries the finding's own time window, so it shows tomorrow what it shows now.

Actions

Sign in (IAP) to acknowledge, snooze or resolve.

Where the numbers come from

sourcelive
collected_atNone
age_secondsNone
staleFalse
shownNone
totalNone
truncatedFalse
coverage_fromNone
retain_daysNone
noteerror-severity job lines from BigQuery, last hour, burst-grouped per job
detectorjob_failing