{ "apiVersion": "monitoring.coreos.com/v1", "kind": "PrometheusRule", "metadata": { "labels": { "app.kubernetes.io/component": "prometheus", "app.kubernetes.io/instance": "k8s", "app.kubernetes.io/name": "prometheus", "app.kubernetes.io/part-of": "kube-prometheus", "app.kubernetes.io/version": "2.44.0", "prometheus": "k8s", "role": "alert-rules" }, "name": "prometheus-k8s-prometheus-rules", "namespace": "monitoring" }, "spec": { "groups": [ { "name": "prometheus", "rules": [ { "alert": "PrometheusBadConfig", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} has failed to reload its configuration.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheusbadconfig", "summary": "Failed Prometheus configuration reload." }, "expr": "# Without max_over_time, failed scrapes could create false negatives, see\n# https://www.robustperception.io/alerting-on-gauges-in-prometheus-2-0 for details.\nmax_over_time(prometheus_config_last_reload_successful{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) == 0\n", "for": "10m", "labels": { "severity": "critical" } }, { "alert": "PrometheusNotificationQueueRunningFull", "annotations": { "description": "Alert notification queue of Prometheus {{$labels.namespace}}/{{$labels.pod}} is running full.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheusnotificationqueuerunningfull", "summary": "Prometheus alert notification queue predicted to run full in less than 30m." }, "expr": "# Without min_over_time, failed scrapes could create false negatives, see\n# https://www.robustperception.io/alerting-on-gauges-in-prometheus-2-0 for details.\n(\n predict_linear(prometheus_notifications_queue_length{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m], 60 * 30)\n>\n min_over_time(prometheus_notifications_queue_capacity{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m])\n)\n", "for": "15m", "labels": { "severity": "warning" } }, { "alert": "PrometheusErrorSendingAlertsToSomeAlertmanagers", "annotations": { "description": "{{ printf \"%.1f\" $value }}% errors while sending alerts from Prometheus {{$labels.namespace}}/{{$labels.pod}} to Alertmanager {{$labels.alertmanager}}.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheuserrorsendingalertstosomealertmanagers", "summary": "Prometheus has encountered more than 1% errors sending alerts to a specific Alertmanager." }, "expr": "(\n rate(prometheus_notifications_errors_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m])\n/\n rate(prometheus_notifications_sent_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m])\n)\n* 100\n> 1\n", "for": "15m", "labels": { "severity": "warning" } }, { "alert": "PrometheusNotConnectedToAlertmanagers", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} is not connected to any Alertmanagers.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheusnotconnectedtoalertmanagers", "summary": "Prometheus is not connected to any Alertmanagers." }, "expr": "# Without max_over_time, failed scrapes could create false negatives, see\n# https://www.robustperception.io/alerting-on-gauges-in-prometheus-2-0 for details.\nmax_over_time(prometheus_notifications_alertmanagers_discovered{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) < 1\n", "for": "10m", "labels": { "severity": "warning" } }, { "alert": "PrometheusTSDBReloadsFailing", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} has detected {{$value | humanize}} reload failures over the last 3h.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheustsdbreloadsfailing", "summary": "Prometheus has issues reloading blocks from disk." }, "expr": "increase(prometheus_tsdb_reloads_failures_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[3h]) > 0\n", "for": "4h", "labels": { "severity": "warning" } }, { "alert": "PrometheusTSDBCompactionsFailing", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} has detected {{$value | humanize}} compaction failures over the last 3h.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheustsdbcompactionsfailing", "summary": "Prometheus has issues compacting blocks." }, "expr": "increase(prometheus_tsdb_compactions_failed_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[3h]) > 0\n", "for": "4h", "labels": { "severity": "warning" } }, { "alert": "PrometheusNotIngestingSamples", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} is not ingesting samples.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheusnotingestingsamples", "summary": "Prometheus is not ingesting samples." }, "expr": "(\n rate(prometheus_tsdb_head_samples_appended_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) <= 0\nand\n (\n sum without(scrape_job) (prometheus_target_metadata_cache_entries{job=\"prometheus-k8s\",namespace=\"monitoring\"}) > 0\n or\n sum without(rule_group) (prometheus_rule_group_rules{job=\"prometheus-k8s\",namespace=\"monitoring\"}) > 0\n )\n)\n", "for": "10m", "labels": { "severity": "warning" } }, { "alert": "PrometheusDuplicateTimestamps", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} is dropping {{ printf \"%.4g\" $value }} samples/s with different values but duplicated timestamp.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheusduplicatetimestamps", "summary": "Prometheus is dropping samples with duplicate timestamps." }, "expr": "rate(prometheus_target_scrapes_sample_duplicate_timestamp_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) > 0\n", "for": "10m", "labels": { "severity": "warning" } }, { "alert": "PrometheusOutOfOrderTimestamps", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} is dropping {{ printf \"%.4g\" $value }} samples/s with timestamps arriving out of order.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheusoutofordertimestamps", "summary": "Prometheus drops samples with out-of-order timestamps." }, "expr": "rate(prometheus_target_scrapes_sample_out_of_order_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) > 0\n", "for": "10m", "labels": { "severity": "warning" } }, { "alert": "PrometheusRemoteStorageFailures", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} failed to send {{ printf \"%.1f\" $value }}% of the samples to {{ $labels.remote_name}}:{{ $labels.url }}", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheusremotestoragefailures", "summary": "Prometheus fails to send samples to remote storage." }, "expr": "(\n (rate(prometheus_remote_storage_failed_samples_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) or rate(prometheus_remote_storage_samples_failed_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]))\n/\n (\n (rate(prometheus_remote_storage_failed_samples_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) or rate(prometheus_remote_storage_samples_failed_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]))\n +\n (rate(prometheus_remote_storage_succeeded_samples_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) or rate(prometheus_remote_storage_samples_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]))\n )\n)\n* 100\n> 1\n", "for": "15m", "labels": { "severity": "critical" } }, { "alert": "PrometheusRemoteWriteBehind", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} remote write is {{ printf \"%.1f\" $value }}s behind for {{ $labels.remote_name}}:{{ $labels.url }}.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheusremotewritebehind", "summary": "Prometheus remote write is behind." }, "expr": "# Without max_over_time, failed scrapes could create false negatives, see\n# https://www.robustperception.io/alerting-on-gauges-in-prometheus-2-0 for details.\n(\n max_over_time(prometheus_remote_storage_highest_timestamp_in_seconds{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m])\n- ignoring(remote_name, url) group_right\n max_over_time(prometheus_remote_storage_queue_highest_sent_timestamp_seconds{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m])\n)\n> 120\n", "for": "15m", "labels": { "severity": "critical" } }, { "alert": "PrometheusRemoteWriteDesiredShards", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} remote write desired shards calculation wants to run {{ $value }} shards for queue {{ $labels.remote_name}}:{{ $labels.url }}, which is more than the max of {{ printf `prometheus_remote_storage_shards_max{instance=\"%s\",job=\"prometheus-k8s\",namespace=\"monitoring\"}` $labels.instance | query | first | value }}.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheusremotewritedesiredshards", "summary": "Prometheus remote write desired shards calculation wants to run more than configured max shards." }, "expr": "# Without max_over_time, failed scrapes could create false negatives, see\n# https://www.robustperception.io/alerting-on-gauges-in-prometheus-2-0 for details.\n(\n max_over_time(prometheus_remote_storage_shards_desired{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m])\n>\n max_over_time(prometheus_remote_storage_shards_max{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m])\n)\n", "for": "15m", "labels": { "severity": "warning" } }, { "alert": "PrometheusRuleFailures", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} has failed to evaluate {{ printf \"%.0f\" $value }} rules in the last 5m.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheusrulefailures", "summary": "Prometheus is failing rule evaluations." }, "expr": "increase(prometheus_rule_evaluation_failures_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) > 0\n", "for": "15m", "labels": { "severity": "critical" } }, { "alert": "PrometheusMissingRuleEvaluations", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} has missed {{ printf \"%.0f\" $value }} rule group evaluations in the last 5m.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheusmissingruleevaluations", "summary": "Prometheus is missing rule evaluations due to slow rule group evaluation." }, "expr": "increase(prometheus_rule_group_iterations_missed_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) > 0\n", "for": "15m", "labels": { "severity": "warning" } }, { "alert": "PrometheusTargetLimitHit", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} has dropped {{ printf \"%.0f\" $value }} targets because the number of targets exceeded the configured target_limit.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheustargetlimithit", "summary": "Prometheus has dropped targets because some scrape configs have exceeded the targets limit." }, "expr": "increase(prometheus_target_scrape_pool_exceeded_target_limit_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) > 0\n", "for": "15m", "labels": { "severity": "warning" } }, { "alert": "PrometheusLabelLimitHit", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} has dropped {{ printf \"%.0f\" $value }} targets because some samples exceeded the configured label_limit, label_name_length_limit or label_value_length_limit.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheuslabellimithit", "summary": "Prometheus has dropped targets because some scrape configs have exceeded the labels limit." }, "expr": "increase(prometheus_target_scrape_pool_exceeded_label_limits_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) > 0\n", "for": "15m", "labels": { "severity": "warning" } }, { "alert": "PrometheusScrapeBodySizeLimitHit", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} has failed {{ printf \"%.0f\" $value }} scrapes in the last 5m because some targets exceeded the configured body_size_limit.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheusscrapebodysizelimithit", "summary": "Prometheus has dropped some targets that exceeded body size limit." }, "expr": "increase(prometheus_target_scrapes_exceeded_body_size_limit_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) > 0\n", "for": "15m", "labels": { "severity": "warning" } }, { "alert": "PrometheusScrapeSampleLimitHit", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} has failed {{ printf \"%.0f\" $value }} scrapes in the last 5m because some targets exceeded the configured sample_limit.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheusscrapesamplelimithit", "summary": "Prometheus has failed scrapes that have exceeded the configured sample limit." }, "expr": "increase(prometheus_target_scrapes_exceeded_sample_limit_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) > 0\n", "for": "15m", "labels": { "severity": "warning" } }, { "alert": "PrometheusTargetSyncFailure", "annotations": { "description": "{{ printf \"%.0f\" $value }} targets in Prometheus {{$labels.namespace}}/{{$labels.pod}} have failed to sync because invalid configuration was supplied.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheustargetsyncfailure", "summary": "Prometheus has failed to sync targets." }, "expr": "increase(prometheus_target_sync_failed_total{job=\"prometheus-k8s\",namespace=\"monitoring\"}[30m]) > 0\n", "for": "5m", "labels": { "severity": "critical" } }, { "alert": "PrometheusHighQueryLoad", "annotations": { "description": "Prometheus {{$labels.namespace}}/{{$labels.pod}} query API has less than 20% available capacity in its query engine for the last 15 minutes.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheushighqueryload", "summary": "Prometheus is reaching its maximum capacity serving concurrent requests." }, "expr": "avg_over_time(prometheus_engine_queries{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) / max_over_time(prometheus_engine_queries_concurrent_max{job=\"prometheus-k8s\",namespace=\"monitoring\"}[5m]) > 0.8\n", "for": "15m", "labels": { "severity": "warning" } }, { "alert": "PrometheusErrorSendingAlertsToAnyAlertmanager", "annotations": { "description": "{{ printf \"%.1f\" $value }}% minimum errors while sending alerts from Prometheus {{$labels.namespace}}/{{$labels.pod}} to any Alertmanager.", "runbook_url": "https://runbooks.prometheus-operator.dev/runbooks/prometheus/prometheuserrorsendingalertstoanyalertmanager", "summary": "Prometheus encounters more than 3% errors sending alerts to any Alertmanager." }, "expr": "min without (alertmanager) (\n rate(prometheus_notifications_errors_total{job=\"prometheus-k8s\",namespace=\"monitoring\",alertmanager!~``}[5m])\n/\n rate(prometheus_notifications_sent_total{job=\"prometheus-k8s\",namespace=\"monitoring\",alertmanager!~``}[5m])\n)\n* 100\n> 3\n", "for": "15m", "labels": { "severity": "critical" } } ] } ] } }