On Friday, I upgraded from Rancher v1.1.4 to v1.2.0 and ran into some serious problems. I was able to resolve these problems (Thanks Vincent!), and wanted to share a tip.
The old instructions for installing Rancher v1.1.4 said to start rancher-server with
$ sudo docker run -d --restart=always --name=rancher-server rancher/server
However, it’s much better to use
--restart=unless-stopped, which is now reflected in the updated v1.1 instructions.
$ sudo docker run -d --restart=unless-stopped -p 8080:8080 rancher/server
If you don’t make this change to Rancher 1.1.4, and then upgrade to Rancher v1.2.0 and then reboot the machine, you may end up with two Rancher Server containers-- one for v1.1.4 and a second one for v1.2.0, and a smattering of duplicate infrastructure containers in your infrastructure (see a screenshot at https://snag.gy/Uxr24e.jpg). This caused major problems for my Rancher cluster, including errors like “Timeout getting IP address” with our Rancher infrastructure containers.