Closed dhoffi closed 4 years ago
That would be these lines (https://github.com/cloudfoundry-community/jumpbox-boshrelease/blob/master/jobs/jumpbox/templates/bin/watcher#L140-L142) overwriting your user/group databases (/etc/passwd
, /etc/group
, /etc/shadow
, and /etc/gshadow
) with cached copies from the first time watcher starts up.
This was implemented on the assumption that people would rebuild their Jumpboxen with new software (provided by BOSH) rather than use system packaging tools to augment (a brittle solution for other reasons).
You are running square into that. One workaround would be to restart watcher job after installing docker:
$ sudo -i
# monit stop watcher
# watch monit summary
# apt --yes install docker-ce
# usermod -a -G docker $USER
# monit start watcher
Although I think this is yet another case of BOSH releases being a bad way to provide jumpbox-y software. I've been exploring a better way, using ephemeral, user-supplied Docker images over in https://github.com/jhunt/containers-boshrelease - particularly the jumpbox
job (https://github.com/jhunt/containers-boshrelease/blob/master/jobs/jumpbox/spec). I'd be interested in your thoughts on that approach.
Hi,
(maybe related to #68 but information there didn't help me solving the problem)
I installed docker on the jumpbox:
well and it works (for a short time)
then all of a sudden and reproducable the docker group gets lost:
waiting 2-5 minutes doin' nothing, really nothing, and then:
I also removed the message-bus line as mentioned in #63 and rebooted but didn't help anything.
any ideas?