Skip to content

Conversation

@gigablah
Copy link
Contributor

Addresses the following issues: #42, #43, #57, #103, #136

This is a rewrite of generateFromEvents that makes the following changes:

  • Only configs that have watch = true will be generated upon container events.
  • All configs will be regenerated when the events channel is reattached, rather than upon docker client re-initialization. (avoids missing out on events in between)
  • Each config with watch now has its own separate channel. This is to enable individual debouncing parameters, specified with the wait parameter. Syntax is min:max, e.g. 500ms:2s. This feature is referenced from hashicorp/envconsul

I've also added an integration test for generator.go that verifies the wait parameter :)

@gigablah
Copy link
Contributor Author

A few more changes:

  • I made generateAtInterval and generateFromEvents aware of termination signals as well (replaced an apparently unused quit channel in generateAtInterval in the process)
  • Made the docker daemon reconnection logic more robust, so now it will survive a docker restart
  • Also added an internal retry flag for the generator which determines if it will attempt to reconnect if the events channel is closed, mainly for testing purposes.

@jgallen23
Copy link

this is awesome! This solves all my issues and would let me safely run docker-gen in prod and give me nearly zero-downtime deployments

jwilder added a commit that referenced this pull request Jan 26, 2016
Add -wait parameter to debounce watched events
@jwilder jwilder merged commit f59341e into nginx-proxy:master Jan 26, 2016
@jwilder
Copy link
Collaborator

jwilder commented Jan 26, 2016

Nice @gigablah! Thanks!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants