MinIO

See Collect MinIO Metrics Using Prometheus

The test environment used public metrics:

bash

MINIO_PROMETHEUS_AUTH_TYPE=public

Production deployments can keep authentication and configure the corresponding Bearer token in Categraf. Do not expose metrics publicly.

Collection Configuration

Both dashboards use cluster, bucket, and node metrics. Scraping only /cluster leaves object, bucket, process, and disk panels empty.

toml

[[instances]]
urls = ["http://127.0.0.1:9000/minio/v2/metrics/cluster"]
labels = { job = "minio", instance = "minio-01:9000" }

[[instances]]
urls = ["http://127.0.0.1:9000/minio/v2/metrics/bucket"]
labels = { job = "minio", instance = "minio-01:9000" }

[[instances]]
urls = ["http://127.0.0.1:9000/minio/v2/metrics/node"]
ignore_metrics = [
  "go_*",
  "minio_audit_*",
  "minio_cluster_*",
  "minio_node_replication_*",
  "minio_node_storage_class_*",
  "minio_notify_*",
  "minio_s3_*",
  "minio_software_*"
]
labels = { job = "minio", instance = "minio-01:9000" }

Keep the same job across all three endpoints. Create a bucket and perform real object uploads, downloads, and deletes before validating the dashboards. Scrape cluster/bucket from the load balancer and scrape the node endpoint on every MinIO node with a unique instance. The ignore list prevents duplicate cluster, S3, and Go series from the node endpoint.

MinIO recommends metrics v3 for new deployments, but these dashboards query v2 minio_* names. Update the dashboard PromQL before switching to /minio/metrics/v3.