miget_ prefix and are available in Grafana dashboards and the Metrics API.
App Metrics
Metrics about your application’s deployment status and HTTP traffic.Replicas
Labels:
namespace, app
HTTP
Labels:
namespace, app, status, method
Instance Metrics
Metrics about individual app instances (pods). Each running instance of your app reports these independently.CPU & Memory
Labels:
namespace, instance, app, resource (for limits/requests: cpu or memory)
Network
Labels:
namespace, instance, app
Disk
Labels:
namespace, instance, app, device
Status
Labels:
namespace, instance, app, phase (for status_phase: Pending, Running, Succeeded, Failed)
Volume Metrics
Metrics about persistent storage attached to your apps.
Labels:
namespace, app, addon, addon_type, volume (the two limit metrics also carry device)
miget_volume_used_bytes and miget_volume_size_bytes are read from the filesystem on the volume, so used space includes everything stored there. For a managed PostgreSQL instance that means tables, indexes, WAL and temporary files. Space left for writes is miget_volume_size_bytes - miget_volume_used_bytes. Both are collected every 30 seconds from a reading that refreshes once a minute.
volume is the volume’s name. For managed PostgreSQL it is <service name>-<instance number>, and an instance that is rebuilt gets a new number, so select on addon (the service name) rather than on volume when you build alerts.
Addon Metrics
PostgreSQL
Labels:
namespace, addon, addon_type, instance, database, role
Cache hit ratio is derived rather than reported: blocks_hit / (blocks_hit + blocks_read). It is undefined until something has been read - a database that has served nothing is not a cache missing every time.
Per-database metrics cover the system databases too. Filter with database!~"postgres|template.*" to count only the databases you created; on a small instance the system ones are a large share of the total size.