Local stacks
A cluster’s Grafana talks to that cluster’s Prometheus and keeps working if Astronomer is down.
ASTRONOMER PRODUCT / OBSERVABILITY & RECOVERY
See every cluster. Rehearse the restore.
Astronomer manages the monitoring, alerting and logging pieces you choose on each cluster, gives the fleet one inbox for alerts, routes events to the systems you already use, and treats backup as something you prove, not something you assume.
Alpha · Backends and retention depend on the topology you choose01 / MONITORING
Metrics per cluster, and for the fleet.#
Each cluster can run its own monitoring stack, managed from Astronomer, with its own Prometheus and optional Grafana. Shared stacks add a fleet view on top: Thanos for metrics, Alertmanager and a fleet Grafana with folders for the management plane and each cluster.
Shared stacks
Thanos, Alertmanager and Grafana as shared families. Fleet Grafana is reached through Astronomer with cluster-scoped queries.
Bring your own
Point Astronomer at existing monitoring endpoints instead, and add dashboard widgets, query widgets and saved views.
A hosted Loki for fleet logs is optional and off by default; a sizer checks the management cluster has room before it is enabled. Retention and capacity depend on the topology you choose.
02 / ALERTING
Thresholds where they fit. Baselines where they don’t.#
Alert rules are defined on each cluster. Threshold rules fire when a value crosses a limit. Anomaly rules learn a rolling baseline per cluster and metric and fire on deviation, with a minimum-sample gate so a new rule does not fire on its first data point.
The fleet inbox collects firing alerts, history, silences and inhibitions, and routes them to Slack, email, PagerDuty, Microsoft Teams or a generic webhook.
- Anomaly defaults: 3 standard deviations over a 24-hour window, at least 50 samples
- Silences for planned work; inhibition to suppress dependent alerts
- Astronomer’s own health alerts ship with runbooks
03 / LOGGING
Ship logs where they belong. Query the ones you can.#
Pipelines are configured on each cluster, with namespace selection and filters, and route to shared destinations. Astronomer is explicit about what it can query and what it only ships.
| Destination | From Astronomer | Notes |
|---|---|---|
| Loki | Query from Astronomer | Tenant-scoped query of the system Loki |
| OpenSearch and Elasticsearch | Query from Astronomer | Adapters with timeouts and response limits |
| Splunk | Query from Astronomer | Credentialed adapter |
| Datadog and CloudWatch | Deep link | Astronomer opens a safe link into the provider |
| S3 and syslog | Shipping only | Logs are delivered; there is no query |
LOGGINGASTRONOMER
Pod logs stream live from the cluster explorer through the agent, separately from pipelines.
04 / INTEGRATIONS
Events go where your teams already look.#
Forward audit events
Syslog over UDP, TCP or TLS, Splunk HEC and NDJSON over HTTPS, in RFC 5424, RFC 3164, CEF or NDJSON format.
Platform events
Outbound webhooks for fleet events, with configurable payloads, retries and delivery tracking.
SMTP and templates
Transactional email through your SMTP server, with editable notification templates.
Support bundles export diagnostics with redaction. Queue, dead-letter and outbox views show background work that is waiting or failed.
05 / BACKUP AND DISASTER RECOVERY
Two kinds of backup, kept apart.#
Member clusters. Install Velero from the catalog, configure S3, MinIO-compatible, GCS or Azure Blob storage, then take snapshots and schedule them from the cluster’s Snapshots page, including CSI volume snapshots. Restores are tracked as runs.
The management plane. Once S3-compatible storage and credentials are configured, a nightly job dumps Astronomer’s PostgreSQL. A weekly drill restores the latest dump into a scratch PostgreSQL, checks the schema and row counts, and records the result. An alert fires if no drill has succeeded in 14 days.
Key custody is part of recovery: the encryption key must match the database you restore. A release claim needs your own restore rehearsal and storage durability evidence.
FAQ
Operations questions.#
Do we have to use Astronomer’s monitoring stack?
No. The automatic baseline only adds kube-state-metrics and the node exporter when Quick Start is selected. Monitoring stacks, Loki, Grafana and logging are choices, and you can point Astronomer at endpoints you already run. #
What happens to dashboards if Astronomer is down?
Cluster Grafana keeps working against its own Prometheus. The fleet Grafana is reached through Astronomer and is unavailable while Astronomer is down. #
Can Astronomer restore a whole cluster?
Astronomer manages Velero snapshots, schedules and restores on member clusters. Recreating the cluster itself is your provisioning tool’s job; Astronomer then adopts it and delivers to it again. #
NEXT STEP
See Astronomer with your environment in mind.
Tell AlphaBravo what you run today. We will show you the workflows that matter to your team.