I'm deploying a Docker app on Elastic Beanstalk. It works, but takes forever (30 minutes until it's back in the "green" state). On my development environments, it finishes within a few seconds. SSHing into the production instance and tailing the logs (/var/log/eb-activity.log), I see that it hangs on the starred line below (running 03build.sh):
[2016-05-23T13:10:00.430Z] INFO [3199] - [Application deployment app-160523_130556@23/StartupStage0/AppDeployPreHook/02loopback-check.sh] : Starting activity...
[2016-05-23T13:10:07.845Z] INFO [3199] - [Application deployment app-160523_130556@23/StartupStage0/AppDeployPreHook/02loopback-check.sh] : Completed activity.
*****[2016-05-23T13:10:07.845Z] INFO [3199] - [Application deployment app-160523_130556@23/StartupStage0/AppDeployPreHook/03build.sh] : Starting activity...
[2016-05-23T13:31:58.805Z] INFO [3199] - [Application deployment app-160523_130556@23/StartupStage0/AppDeployPreHook/03build.sh] : Completed activity. Result:
latest: Pulling from srfoster/my-repo
6714a7dc486b: Pulling fs layer
5866500c2af9: Pulling fs layer
... ETC ...
Based on one of the (not accepted) answers to this similar question...
Why are Docker build commands running so slow in Elastic Beanstalk?
I ran
docker info | grep Storage
and got back
devicemapper
Supposedly, aufs is faster than devicemapper (according to the question linked above).
Buuuut, even if that's true, I wouldn't know how to make Beanstalk's prebuilt images use one versus the other when spinning up.
Any suggestions on this? Or is there some other way to go about fixing the slow deployments?