[zacharyvoase]

I’m really sorry to get all critical, but if you personally have to SSH into the server to do setup or deployment—especially if you have to be root—you’re doing it wrong.

I also feel that if anyone or anything has to SSH into a server for deployment, it could be done better. I use Chef to automate all of this stuff; it's surprising that we have great tools like Chef, Puppet and CFEngine and people still feel the need to write custom collections of fragile scripts to get basic stuff done.

[ajvb]

Cool story bro.

Did you look at the actual main aspect of this simple tool, or just the one folder with the server-script? As previously mentioned. The main aspect of this is the automatic _Django Project Builder_, not "a collection of fragile scripts to get basic stuff done".

[meric]

Of course deploying as root is doing it wrong.

However, I still haven't found enough reason to switch from using SSH for deployment. It's almost always three lines. ssh; git pull; ./manage.py syncdb/migrate/collectstatic.

The only hindrance has been configuring the site to work on the server for the first time, i.e what the site at this link claims to solve.

I've spent several hours on fabric before but gave up when I realised I've spent more time on learning it than the time I spent deploying my code.

Can you tell me what I am missing out? I'm still relatively new to django deployment and I feel I'm missing something but I haven't found it yet.

[heretohelp]

Idempotence, edge-cases, compliance ensurance.

What if you want to redeploy a configuration change to your web servers? What if you want to ensure such changes get re-deployed everytime a change is made to a config file?

[meric]

I guess if you are deploying the same site to more than 1 production server these will become worthy problems to solve?

[heretohelp]

These are problems that can be solved in an afternoon, and then never repeated ever again.

You need to be using automation.

We have something like 16-17 servers, of which 4-5 are actually production servers. Every single one of them was provisioned, configured, and are pushed to via my code.

Everything from haproxy, to the frontend web servers, to the backend web servers, to the frontend assets, the CDN, the image server, the backups, our staging cluster, experimental cluster, databases...all of it is automated.

Do it once, do it right, do it in code, never do it manually again.

The idempotence is just so you can repeat the same job over and over if it fails halfway between without causing any breakage.

[whimsy]

You're right, I should be using these. How's the learning curve? Should I find a tutorial? Which should I choose, and why?