[
https://issues.apache.org/jira/browse/ZOOKEEPER-1940?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14031328#comment-14031328
]
David Medinets commented on ZOOKEEPER-1940:
-------------------------------------------
~rtice - I looked at the signalfuse project but was unable to make it work. I
haven't studied Python and that project did not provide batch files to simplify
startup. Naively building and running their image resulted in the following
message. The maestro-ng approach might be better than using pipework, but I
wasn't able to use the project with my current knowledge.
{noformat}
$ docker run -i -t docker/zookeeper:3.4.5
Traceback (most recent call last):
File "/opt/zookeeper-3.4.5/.docker/run.py", line 25, in <module>
ZOOKEEPER_NODE_LIST = get_node_list(get_service_name(),
File "/usr/local/lib/python2.7/dist-packages/maestro/guestutils.py", line 27,
in get_service_name
raise MaestroEnvironmentError('Service name was not defined')
maestro.guestutils.MaestroEnvironmentError: Service name was not defined
{noformat}
> Integrate with Docker.
> ----------------------
>
> Key: ZOOKEEPER-1940
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-1940
> Project: ZooKeeper
> Issue Type: Wish
> Reporter: David Medinets
> Priority: Trivial
>
> Docker is an open platform for developers and sysadmins to build, ship, and
> run distributed applications.
> It's become quite popular and I'd like to see the zookeeper community suggest
> a standard way to run zookeeper inside docker containers. To get the
> conversation started, I have a working example at:
> https://github.com/medined/docker-zookeeper
> I hope there is a better technique that I used. And if there is please make
> suggestions.
> The difficulty, I think, posed by Docker, is that the images are started
> before the bridge network is created. This means, again I think, that
> zookeeper is running inside the container with no way to communicate with the
> ensemble for some non-trivial amount of time.
> My resolution to this was to force each each to wait 30 seconds before
> starting zookeeper. I still see connection errors in the logs, but eventually
> the cluster settles and everything seems to work.
> I'm hoping that someone which more networking experience than I can find a
> way to eliminate that 30 second delay and the connection errors during
> startup.
> Thanks for reading this far.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
