clusterhq / docker-images Goto Github PK

what are the docker run commands for these 3 (w/o using flocker)
I realize that elasticsearch needs a volume mount point too?

clusterhq/elasticsearch
clusterhq/logstash
clusterhq/kibana

The instructions indicate one should download the latest image from mongodb.org, but then there's a checked in tarball of some sort which is what the script actually uses. Besides inconsistency of instructions, we shouldn't be checking in build artifacts or tarballs, just recipes.

This happens when you move a postgres container from one node to another in a Flocker clsuter.

1. A new password is assigned, rather than reusing the old one.
2. If you already have data in /data/pgsql the container won't start up because it's trying to re-initialize the database unnecessarily:

Aug 11 09:50:00 node2 docker[16809]: PG_PASSWORD=24a3bb1abd
Aug 11 09:50:00 node2 docker[16809]: initdb: directory "/data/pgsql" exists but is not empty
Aug 11 09:50:00 node2 docker[16809]: If you want to create a new database system, either remove or empty
Aug 11 09:50:00 node2 docker[16809]: the directory "/data/pgsql" or run initdb
Aug 11 09:50:00 node2 docker[16809]: with an argument other than "/data/pgsql".

I have no idea where this tarball came from or how to rebuild it. We should have recipes that build the tarballs, rather than opaque binary blobs that are end results, so that we can reproduce the process (and fix things when they break).

I'm pretty sure a single step Dockerfile should suffice, and clusterhq/postgreslite is unnecessary.

I would expect a single step Dockerfile should be able to build the image, with no need for clusterhq/mongolite to exist.