On Thu, 2015-04-30 at 23:18 +0200, Václav Pavlín wrote:
Hello,
I wanted to use CentOS Atomic Host and Docker images to package
atomicapp-run [1] into an image but I ran into troubles with docker and
docker-py (probably).
See http://fpaste.org/217373/30426738/
There is f.e. ${CONFDIR} specified in LABEL INSTALL, but it disappears
during docker build. I "fixed" this with adding a backslash in front of
dollar sign (\${CONFDIR}) so it persisted after the build. But when I ran
atomic install vpavlin/atomicapp-run
I got this incorrect cmd again:
docker run --rm -it --privileged -v /run:/run -v :/atomicapp -v /:/host
-v /answers.conf:/application-entity/answers.conf -e
IMAGE=vpavlin/wp-app -e NAME=wp-app --name wp-app vpavlin/wp-app
containerapp -v install --update --path /atomicapp /application-entity
This works completely fine on RHELAH, but there is a different docker
verison:
RHELAH:
bash-4.2# docker --version
Docker version 1.5.0-dev, build fc0329b/1.5.0
CentOS AH:
bash-4.2# docker --version
Docker version 1.6.0, build bdbc177
So my guess it that docker somehow ignores/evaluates variables in LABELS
in 1.6.0...and even if they persist with the backslash fix, they get
lost by inspect() from docker-py used in atomic tool.
Any thoughts if this is intended or how to fix it/workaround it?
Thanks,
Vašek
[1] https://github.com/vpavlin/atomicapp-run/blob/master/Dockerfile.centos
The patch by Dan was slightly altered before it went live:
https://github.com/docker/docker/commit/6784a772baaa20936f00b8adc00a4b4444356181
How does the inspect output look like after you've escaped the dollars?