Postgres as NONUSE? about ansible-galaxy-extras

[bgruening]

@tiagoantao you are referring to the Galaxy-Docker NONUSE environment variable? If so this would be possible, at least as runtime option. We can simple not start the daemon. Is this what you are looking for?

[tiagoantao]

That is what I have, indeed I change the supervisor configuration file. But I was thinking in a more general solution. Though to be quite frank, I need to think a bit more on this myself.

In my case I have a postgres server outside and LDAP authentication, which means that I tweak the container quite a lot. I have actually thought on fully separating the services (say proftpd - which I also tweak to make it SFTP) and then composing a few containers. But due to lack of time, this never got out of thinking stage. The current solution is great, but a bit monolithic.

[bgruening]

We are thinking a lot about composing the containers and not having a static one. This would be great, but also means we can not use large parts of the ansible playbooks and diverging a little bit from all the other deployments.
The current plan is to move with every new release one service out and see how things are going and how much more complicated the maintenance and the experience is.

Are you interested in up-streaming your sftp changes? Would love to have this in as well.

[tiagoantao]

I am going to test the SFTP changes on Monday. I will do that as soon as they are working.

My long view would be to have a service per container and sharing via volumes. For now I really need to separate postgres (which I did).
Another thing is plugable authentication schemes. In my case I use LDAP and I had to patch proftp and galaxy.

As I am going grossly off-topic here is another one: having an integration with the sequencer. I am working on having a samba server (required by illumina) dropping directly on a galaxy "FTP" drop.

[mvdbeek]

We are thinking a lot about composing the containers and not having a static one. This would be great, but also means we can not use large parts of the ansible playbooks and diverging a little bit from all the other deployments.

Actually I think that the playbooks/roles are awesome for this, they just need more tags so you can skip/include the parts you want. Well, and there is still issue #1, which would help as well.
Maybe we can plan this out at the GCC hackathon.

[bgruening]

@mvdbeek and this is exactly what I said. We will not use (skip) large parts of the playbooks integrate a bunch of new code-paths that are less tested, making roles more and more complex.
I'm not against this, we do have plans to make it more modular but I see also downsides and we need to have a nice plan in advance. In the end all this needs to be maintained :)