Logs

Logs

Get logs for an app. By default, convox logs streams logs continuously. Use --no-follow to print current logs and exit.

Usage

    convox logs

Examples

    $ convox logs
    2026-03-18T12:47:41Z service/web/77f0e67e-4886-4aa8-be56-1d19a3aab53b ns=template id=2bdd60aaf431 route=root at=end state=success elapsed=0.065
    2026-03-18T12:47:41Z service/web/77f0e67e-4886-4aa8-be56-1d19a3aab53b ns=template id=2bdd60aaf431 route=root at=start method="GET" path="/" elapsed=0.029
    2026-03-18T12:47:43Z service/web/a81ba08c-6dbe-48a4-88e6-da5f940156ae ns=template id=57c9464c88f6 route=root at=end state=success elapsed=0.070
    2026-03-18T12:47:43Z service/web/77f0e67e-4886-4aa8-be56-1d19a3aab53b ns=template id=f5b0fcdd6f63 route=root at=start method="GET" path="/" elapsed=0.038
    ....

    $ convox logs --filter 2bdd60aaf431 --since 24h
    2026-03-18T12:47:41Z service/web/77f0e67e-4886-4aa8-be56-1d19a3aab53b ns=template id=2bdd60aaf431 route=root at=end state=success elapsed=0.065
    2026-03-18T12:47:41Z service/web/77f0e67e-4886-4aa8-be56-1d19a3aab53b ns=template id=2bdd60aaf431 route=root at=start method="GET" path="/" elapsed=0.029

Flags

Flag	Short	Description
--filter		Filter for a specific string within the logs
--no-follow		Print logs and exit rather than streaming
--since		Time frame for log query (e.g., 24h, 2m; default 2m)
--service	-s	Filter to a specific service
--tail		With --service, the number of lines to read from the end of each Process's log. Ignored without --service
--allow-previous		Include logs from previous container instances
--max-log-requests		Maximum number of concurrent log follow streams (default 20)

Service Logs

convox logs --service <name> reads each of the Service's Processes directly from Kubernetes. On Rack version 3.25.11 or later it returns every line each Process wrote in the --since window (default 2m), in follow and --no-follow mode. Earlier Racks return the last 10 lines of each Process unless --tail is set.

Command	Returns
convox logs -s web	each Process's lines from the last 2 minutes, then the live stream
convox logs -s web --since 2h --no-follow	each Process's lines from the last 2 hours, then exits
convox logs -s web --tail 10	the last 10 lines of each Process within the last 2 minutes, then the live stream
convox logs -s web --since 0s --no-follow	each Process's whole current log file, then exits

    $ convox logs -s web --since 2h --no-follow
    [pod/web-7cd6cc78dd-blvtn/myapp] 2026-03-18T12:47:41.364333569Z GET / 200
    [pod/web-7cd6cc78dd-blvtn/myapp] 2026-03-18T12:47:43.108422907Z GET /other 404
    [pod/web-7cd6cc78dd-x2kq9/myapp] 2026-03-18T12:47:42.920115381Z GET / 200
    [pod/web-7cd6cc78dd-x2kq9/myapp] 2026-03-18T12:47:44.511290637Z GET /health 200

• Each line starts with [pod/<process>/<app>] and its timestamp, unlike the service/<service>/<process> prefix of the whole-App view.
• The read covers the current log file of the Service's current Processes. A Process replaced inside the window is absent, and log rotation on the node can drop the oldest part of a long window.
• With --no-follow and more than one Process, each Process's lines print as one block, not merged by time. In follow mode the streams interleave.
• --since 0s reads the whole log file only with --service. Without --service, 0s starts the read at the current time and returns no earlier lines.
• --filter is ignored with --service.

Tailing High-Concurrency Services

convox logs --service <name> streams logs from every pod that matches the selector. The underlying kubectl plumbing caps concurrency at 20 follow streams by default. Following the logs of a Service that runs more than 20 Processes fails with:

ERROR: you are attempting to follow 50 log streams, but maximum allowed concurrency is 20, use --max-log-requests to increase the limit

A --no-follow read has no cap and reads each Process in turn. Pass --max-log-requests N to raise the concurrency cap when following.

    $ convox logs --service pii-worker --since 5m --max-log-requests 50

Set the value to at least the number of Processes the Service runs. Log streams are persistent HTTP connections, so very large values put sustained load on the API server.

Log Sources on AWS Racks

On an AWS Rack the whole-App view and the per-Service view come from different sources. convox logs without --service is served from the App's CloudWatch log group, while convox logs --service <name> reads Pod logs directly. CloudWatch applies the --filter pattern, so --filter narrows the whole-App view and is ignored when --service is set. On a Rack with cloudwatch_disable set to true, the whole-App view returns empty output while --service continues to work.

Log Sources on GCP, Azure, DigitalOcean and Metal Racks

On GCP, Azure, DigitalOcean and metal Racks, convox logs without --service reads the Rack's Elasticsearch, and convox logs --service <name> reads Pod logs directly. From Rack version 3.25.11, --no-follow prints every line in the --since window, oldest first. The live stream prints each line once. A line whose timestamp is up to 5 seconds older than the newest line already printed still prints when it arrives, so the live output is not strictly in timestamp order. A line more than 5 seconds older than that does not appear in the live stream; --no-follow shows it. --tail applies only with --service. See Logging.

See Also

• Logging for log configuration and forwarding
• deploy-debug for diagnosing pods that never reach a ready state