1180767 - migrate oneanddone from PaaS to python cluster

Status: RESOLVED WONTFIX

(Infrastructure & Operations Graveyard :: WebOps: Other, task)

Description

[Jake Maul [:jakem]]

As a fix for PaaS trouble, we're just going to migrate this to our Python cluster.

AWS/Nubis would also be a reasonable fit, but there aren't resources available to help make that work... everything there is dedicated to bedrock, bugzilla, vidyo, ldap, and okta (I think those are the 5, anyway).

Comment 1

[Jake Maul [:jakem]]

oneanddone-dev.allizom.org is alive and working.

@mkelly: want to test and let us know if everything seems okay? I haven't set up any crons or migrated data, or done anything with chief/updating...

Comment 2

[Jake Maul [:jakem]]

Note:

oneanddone.allizom.org (stage env) is also set up, but as that name already goes to PaaS I haven't changed DNS for it. You can access it with this in your /etc/hosts:

63.245.217.30 oneanddone.allizom.org

Comment 3

[Bob Silverberg [:bsilverberg]]

Jake, I was able to get to the home page, but logging in doesn't work, and with there being no data in the database it's hard to test anything else. The error emails that are set up on the paas don't seem to be working as usually, if there is a problem logging in, I would receive an email showing me the error.

I suggest we migrate the data from the current dev database into a database that the new app can access (or maybe we can just point the new app to the existing dev database?), which will allow for much better testing/verification. I also need an admin user set up so I can access the admin. If I can ssh into the environment in which the app is running I can probably set that up for myself.

As I am not getting error emails, we need to either fix that, or let me know how I can see errors some other way (e.g., via logs on the server). I apologize that I do not have any experience working with sites on the python cluster, so I don't know what sort of information might be available to me for debugging, and therefore I'll need a bit of help with that. I also don't know anything about chief, but am a quick learner. I assume there is some facility that will allow us to specify settings for the different environments much the same way we did using environment variables on Stackato.

Mike isn't really on the One and Done project anymore, and while I would more than welcome his help I'm guessing he's busy with other things, so it's probably best to needinfo me when things need to get done.

Let me know what else I can do at this point, and thanks for getting things this far!

Comment 4

[Bob Silverberg [:bsilverberg]]

I was also able to access staging after updating my hosts file and see the same thing as dev. The site looks okay, but I cannot log in. There could be a number of reasons for this, but one thing I recall that often tripped me up locally when I started working with the site is that if caching is defined to Django, but not running, then log in doesn't work, so that might be one thing to check.

Comment 5

[Bob Silverberg [:bsilverberg]]

Jake, please let me know what, if anything, I can do to help move this forward. Thanks!

Comment 6

[Bob Silverberg [:bsilverberg]]

Jake, I had a quick chat with Mike Kelley to find out how I might get some more debugging info on what's going on on the cluster. He suggested the following:

1. Get ssh access and log in and try running djago shell.
2. Make sure I am getting error emails.

Here are some requests based on this:

1. Can you please grant me ssh access, as per item 1?
2. I had already mentioned item 2 earlier in this bug. On the paas I am getting emails, but not from the cluster, so we need to make sure that is configured properly. How exactly are the local settings set up on the cluster, and is there a way for me to look at and modify them?
3. Can we get the app pointed at the database that dev is using on the paas, or migrate that data over? There is not much testing I can do without data?
4. If we complete 3, then I will also have admin access. If for some reason completing 3 in the near term is not possible, could I at least get admin access to django so I can create some data of my own?

Comment 7

[Bob Silverberg [:bsilverberg]]

We have decided to pursue another avenue, as getting the site up in the data center is only a stop-gap measure anyway. After some research and experimentation it looks like we can get the site up on Heroku without too much pain and effort, so that is what we plan to do.

Thanks for your help with this.