So Docker 1.13 is out, and it lists a whole host of improvements on it’s release notes page, including:
- better Swarm support including docker-compose modeling – so you can build your Swarm with one file.
- better cleanup commands – so that the mess docker leaves behind can be more easily removed
- changes to the CLI – to improve some command models
- better logging support – mostly for multi-node/Swarm deployments, discovering where a service is/was
This all looks great, so of course I tried re-building a few services I have that leverage docker 1.13. Both models (a Kubernetes deployment and an OpenStack Kolla deployment) blew up. And while I went looking through release notes to figure out what it is that might have caused this, I still haven’t hit on the actual issue in either case.
So, I went back to Docker 1.12, and everything works again….
This for me is the last straw. I’m a huge fan of Container-based deployments, and the Dockerfile build model really seems to be a slick way of building out Container images that are both efficient, and flexible. But I can’t work with a tool where a dot release breaks things in un-fathomable ways. Who has time to debug these sorts of issues when working with production environments, which this release of Docker should support. This is a production release right?
So I’m going to dive into rkt as an alternative to the Docker deployments where I can (thanks CoreOS for getting some support into Kubernetes already!) and will deal with an older version of Docker where I can’t (Kolla team, perhaps this is where I have to jump in with both feet).
I really hope that Docker can get their act together, but I’m not hopeful given past performance.
Founder and CTO