Introducing ClusterFsck: Filebased S3 Configuration Kontrol

briefs.amicushq.com

42 points·tobowers··11 comments

Open articleView on HN

More options is good! Centrally storing configuration is always a good idea.

Amazon OpsWorks manages configuration centrally via a web-based command and control.

Heroku manages configuration centrally via environment and heroku config:add.

ClusterFsck (did I type that right? :)) lets me store things on S3, centrally.

I can't read the name other than ClusterFuck. Sorry, it's my brain. :)

Feature not a bug!

I guess I'm a prude but the name alone would prevent me from using it.

Is this based on Amazon S3 because you have that at hand, or is it based on it because of some innate advantage?

A little bit of both I guess? It's become standard enough that most web developers (at least most I know) have an account and use it for a few things here and there, it's dirt cheap and widely available, and it's just text files we're storing, so nothing more heavy weight is needed. It does also offer versioning, which I believe we may add a feature to take advantage of for rolling back configuration changes in a coherent way if something goes wrong or to recreate a problem locally on a certain date, but don't think we've started even planning that yet (pull requests welcome!)

And, importantly, has ACL controls built in so it's easy to add/remove access to team members.

Don't forget that its highly durable. When was the last time S3 was down?

Have you had any issues with eventual consistency? Does this need to be run on US-west or EU for read-after-write consistency?

Amazon is not involved in Prism, it seems. So that helps.

shameless plug, speaking of S3, I created a DSL for Amazon S3 (using s3cmd, because it is intended for my rake tasks) - http://jjuliano.github.io/s3cmd-dsl

Reply on news.ycombinator.com