Observability
Putnami services are observable by construction: the runtime emits structured logs, metrics, and traces by default, in the same shape locally, in preview, and in production. Debugging starts from system state, not guesswork.
Telemetry is exported over OTLP. The platform provisions a workspace collector that exports to that workspace's runtime project and makes the signals queryable through the same workspace boundary.
The shape
| Signal | Emitted by default |
|---|---|
| Logs | structured, correlated to a request and revision |
| Metrics | runtime and request metrics over OTLP |
| Traces | per-request spans over OTLP |
Because every running revision is attributable to a commit, a signal can be traced back to the exact code that produced it.
Read a module in Live
In putnami.app, open your workspace's Live tab and a module. Its module sheet shows what the served commit declares: HTTP routes and their public access, pages and their actions, gRPC services and methods, databases and schemas, buckets, published and subscribed events, scheduled jobs, needed secret names, configuration keys, migrations, health checks and service calls. CPU, memory and instance limits come from the deployed release.
Live's type filters use these facts. Choose a module's links to highlight its callers and callees and see shared event topics. Links are declared, not observed in production, and the view names the commit they describe.
An unreadable or absent declaration says it was not collected. Sheets need an
activated Source declaration index and an indexed served commit. Pages come
from the committed HTTP route inventory and gRPC services from the committed
gRPC service inventory. A commit indexed before Source read those inventories
shows them as not collected until Source reads it again. A collected page
establishes the web type; a web tag does not establish that a module
serves pages. A gRPC service adds no type.
Query your service
| Want to… | Do this |
|---|---|
| Query recent errors | putnami cloud logs <app> --since 1h --level error |
| Read the logs since the last deploy | putnami cloud logs <app> --since deploy:last |
| Live-tail logs | putnami cloud logs <app> --follow --output=jsonl |
| Query metric series | putnami cloud metrics <app> --window 1h --output=jsonl |
| Find slow traces | putnami cloud traces <app> --slow 500ms --output=jsonl |
All three queries are scoped to the linked workspace and environment. Cursors
keep paged reads bounded; logs --follow is the explicit SSE live-tail mode.
--since deploy:last starts the window when the last deploy that made the app
serve in the environment began, whether a channel move or an operator request
opened it. The window therefore holds the first logs of the
revision that deploy rolled, after the last logs of the revision before. A
deploy that failed before its revision took the traffic does not count.
The deploy is read from the 100 most recent runs of the environment. When none of them made the app serve there, the window starts at the app's last successful release, whichever environment it targeted.
When the deploy history is briefly unavailable, for example while
observability-api starts, --since deploy:last fails with a retryable 503. It
never starts the window at a wrong deploy.