Deployments · drift · activities
Administer your fleet
Every installation you operate as an ordinary mesh node — host, namespace, database, update policy — version-controlled through GitSync and readable in one table.
What ships
- Fleet Console — the app. Every instance on one page: recorded state, the version it is actually RUNNING, sampled health with the sample's age, and exact log links — attention first. Plus the record catalog and the request form for a new instance. An operator surface: it reports the fleet, it does not command it.
- Deployment — the git-backed record: host, namespace, database, update policy. Its Overview is the fleet table; its Logs area is that deployment's slice of the ingested window.
- DeploymentStatus — observation: replicas, restarts, the image actually RUNNING, health — sampled from kube-state-metrics by the
refresh-statusscript. Its Board area is the fleet health board, compared against the latest available image. - LogEntry — a window of Loki, landed as nodes so logs are searchable by meaning. Ingested by
ingest-logs, browsable via Recent. - Skills —
/deploymentto record and inspect,/deployment-activityfor the pattern that keeps provisioning inside the mesh.
Records, not secrets
Deployment nodes are designed to sync to git, so they carry identifiers and topology only. Key Vault fields name secrets; they never hold them.