1145401 - split webapp-django out of socorro

Status: RESOLVED INVALID

(Socorro :: Webapp, task)

Description

[Robert Helmer [:rhelmer]]

Let's split webapp-django out to its own repo. Maybe named "socorro-webapp"?

It'll need its own travis-ci for PRs, and its own package builder.

One of the goals here is to make it possible to push crash-stats changes separately from collection/processing, which I think will naturally fall out of the above.

Comment 1

[[DEACTIVATED] Adrian Gaudebert]

Shouldn't we wait for the removal of the middleware to be complete before doing this? Also, from what I understand, ./webapp-django now depends on files in ./socorro (implementation files of the middleware) so it wouldn't be as straight forward as just putting the webapp-django folder in a new repo.

Comment 2

[Peter Bengtsson [:peterbe]]

Yeah, webapp and socorro.external is going to get very intimate. 

We could of course make socorro a dependency but that's going to either be a heck of a lots of pypi pushing, or using a github sha url in webapp's requirements.txt or having to go back to git submodules. 

My inclination would be to focus first on the extraction of collector and extraction of processor and maybe extraction of crontabber execution.

Comment 3

[Robert Helmer [:rhelmer]]

(In reply to Peter Bengtsson [:peterbe] from comment #2)
> Yeah, webapp and socorro.external is going to get very intimate. 
> 
> We could of course make socorro a dependency but that's going to either be a
> heck of a lots of pypi pushing, or using a github sha url in webapp's
> requirements.txt or having to go back to git submodules. 


We could automate pypi pushing. I think a github sha url would be fine if we want
to do that (let's not go back to submodules :) )


> My inclination would be to focus first on the extraction of collector and
> extraction of processor and maybe extraction of crontabber execution.


I don't think any of these would be too hard. webapp-django is already very self-contained so I think it'd be a bit easier.

collector, processor and crontabber will also have this same problem of depending on a bunch of common code like socorro.external, I don't think this particular problem gets easier with this does it?

Comment 4

[Peter Bengtsson [:peterbe]]

After some discussion in IRC we concluded that 

1) We'll delay splitting out the webapp till last. Doing collector and/or processor first. The reason being that webapp+socorro is going to be rapidly changing due to the dataservice integration lead by lars. 

2) When we do split, lets put in a Github sha to socorro (the base package) in webapp's requirements.txt

3) Once things have really settled (and assert that we have convenient auto-deploy to pypi in socorro) we'll point to a pypi version of socorro in webapp's requirements.txt

No git submodules! :)

Comment 5

[Peter Bengtsson [:peterbe]]

If we're going to split anything out, it'd be the collector or the processors. The bulk of the socorro repo is stuff related to the webapp. E.g. much of the socorro/external/(postgres|es)/ and almost all of the crontabber stuff is to serve the webapp ultimately.