IncidentWebmasterProviderDown

Severity: warning · Fires after: 13h · Rule: central/prometheus/rules/visibility.yml

What it means

One search engine's poll (engine label: google / bing / yandex) has failed three times running. That engine's numbers on the search-visibility dashboard are frozen at the last good poll; the other engines are unaffected.

Expression

webmaster_exporter_up == 0

Polls are 6h apart (VIS_POLL_INTERVAL), so 13h spans three. An engine with no credentials also reads 0 by design (per-engine graceful degradation) — all three are configured, so today 0 means broken. If an engine is ever deliberately unconfigured, exclude it in the rule.

First checks

ssh jakefear@docker2 'cd /opt/jakemon && docker compose logs --since 24h visibility-exporter | grep -iE "poll failed|not configured"'
bin/visibility report --engine <engine> --site wiki.wikantik.com

How to clear

• google — almost always the OAuth refresh token (GOOGLE_REFRESH_TOKEN): revoked, or expired because the OAuth app is in "testing" mode. Re-mint it.
• bing — BING_API_KEY rotated.
• yandex — YANDEX_OAUTH_TOKEN expired (they are time-limited).

Update the root .env, bin/deploy-central.sh, then docker compose up -d --force-recreate visibility-exporter.

Notes

Related: IncidentWebmasterProviderStale — the engine reads healthy but nothing is refreshing.