Right now we update our staging sites on a cron every 5 minutes that tries to pull new data. That's so 90s. Github can ping a URL every time an update is pushed - we should be using that. Oremj set this up on landfill.addons.mozilla.org and it is working well. I'm filing this now because people are having problems with performance and restarts (bug 645739) and this would mean we only update/restart when it's useful. Appparently there are some complications with the firewalls and method of pushing code to servers, but I'm sure we can figure out how to solve that. If it's an in-person thing, we could all talk about it at the all-hands. If I can help, let me know. Wil
A queue seems like a good way to get around firewalls. Github can ping something public which adds a job to the queue and mradm can pull jobs out of the queue.
Can we talk about this during our web push discussions next week? I think it is a great idea but I'd like to see this happen across the board. Problems arise when each site is treated differently and we don't have a standard way of doing things. Systems team will be discussing web pushes internally on wednesday, and pulling anyone in webdev who is interested in on thursday morning to discuss more.
Looking forward to it. thanks
Next and preview will update on post-commit hooks now.
That's awesome. I'm sure Krupa will be pleased.