Remove postgression

[zbynekwinkler]

Based on experience in #1082 I suggest removing postgression. Running tests with it is unusably slow and only brings confusion for the novice wanting to start with gittip.

[tshepang]

Agreed, especially since it's not very clear when one is using that or a local DB. I struggled recently wondering why my make schema took so looong, and resulted in no schema on a local DB, only to discover that there was a remote DB entry in my local.env (which likely got added because I ran make db by mistake). To make up for it, we can just make the docs on creating a local DB easier... it should not be too much to ask for for anyone running a local instance if Gittip to create a local DB too.

[seanlinsley]

It seems we're using Postgression under the theory that it lowers the barrier to entry. Maybe Vagrant is a better option?

[zbynekwinkler]

Shall I prepare pull request? We should decide either way.

[chadwhitacre]

I have anecdotal evidence that Postgression, when it works, does lower the barrier to entry. The intended purpose is to enable someone to get Gittip installed locally as quickly as possible. I've actually seen a guy at a hack night be like, "Whoa, how is Gittip running on my laptop already?" For someone who wants to make tweaks to the Gittip front end, this may be all they need. For someone looking to modify the backend, of course they'll need to configure Postgres locally. I'm not ready to get rid of it. I think we should explain better in our dev docs (which we have yet to really write) what's going on. Leading back into #1272.

[zbynekwinkler]

@seanlinsley @clone1018 @seanlinsley 👍 from #1873 so I am reopening this for further discussion.

Right now postgression works but it has been down for a really long time which actually creates higher barrier to entry than not having it at all. Also http://www.postgression.com/ says that one db instance lasts for 30 minutes. I am running the test with postgression right now to get the actual timing but I am pretty sure it will be over 30 minutes.

[zbynekwinkler]

Well, 30 minutes have passed and the tests started to fail. I got to test_elsewhere_bitbucket.py. As a reference, the whole test suite takes 50s to run on my local machine.

[seanlinsley]

Once we're sure Windows users can use Vagrant without issue (#1887) and we document how to get running with Vagrant (#1846), I think we should stop using Postgression.

[zbynekwinkler]

With recent commits postgression is no longer the thing new developers are pushed into by default but it is left there as an option (which I think given the above is not worth it). I am not sure how to make more sure Windows users can use Vagrant without issue. I have Vagrant running, configured using the script at #1999. I now believe there is one more Vagrant user (right @joeyespo 😉?).

[patcon]

+1 for removing

Has possible security implications, as it assumes the db has zero sensitive information, but we're putting oauth access_tokens into it:
#2239 (comment)

[chadwhitacre]

@patcon Ftr we only started putting oauth access_tokens into it with Venmo.

[chadwhitacre]

(iinm)

[Changaco]

Postgression was removed by @clone1018 in 99bb087.