TallyScrapeJobMissing

absent(up{job="reporting-api"}) or absent(up{job="otel-collector"}) or absent(up{job="openstack-collector"}), for: 5m, in scrape-rules.yaml, which an overlay replaces together with its scrape config. The expression above is the base's.

Symptom

A discovered scrape job resolves to no target at all. up is a per-target series, so zero targets produce no up series rather than up == 0: the job disappears from /targets instead of turning red, and TallyScrapeTargetDown stays silent.

Impact on billing

The same as the job's targets being down, described in TallyScrapeTargetDown, without a red target to see it by.

First checks

1. The Deployment's replica count. A Deployment scaled to zero takes its job off the page.
2. The Service name and the port name the relabel rule keeps, reporting-api;http, otel-collector;metrics and openstack-collector;http in scrape.yaml. A rename on either side drops every endpoint of the job.
3. The victoriametrics-scrape RoleBinding in victoriametrics.yaml. Without it the endpointslice discovery is refused and the job has nothing to keep.
4. The VictoriaMetrics log for discovery errors, an unreachable API server among them.
5. When the firing job is one the deployment does not scrape, compare the overlay's scrape-rules.yaml with its scrape.yaml. A clause left over from the base names a job the overlay never configured, and it fires for as long as the cluster runs. The overlay's file carries one absent clause per job of its scrape.yaml that has kubernetes_sd_configs, and nothing else.