[20160419004044] Log Delivery and Tail Reliability


  • convox/agent:0.67
    • Move to golang:1.6-alpine (3.3) base image. It’s smaller in general and now layers are shared with the Rack API image
    • Write to CloudWatch logs with much simpler format: web:RXZMCQEPDKO/1d11a78279e0 Hello from Docker.
    • Emit custom metric for how many Docker logs lines were read. This offers new operational visibility into the log pipeline:

App Runtime

  • Inject APP, RACK, and AWS_REGION into app process environment just like PROCESS and RELEASE. Re-deploy an app to pick these variables up.


  • convox logs now streams from CloudWatch Logs which is more reliable and easier to use
  • convox logs has new options. Try --since 2h, --follow false and --filter /auth

Roadmap Notes

Now that we’re tailing from CloudWatch Logs, Kinesis isn’t really needed. You can expect more updates that stop using then delete Kinesis streams from every rack and app. It’ll come back as something you can optionally add with convox services add kinesis