Automating deployments to the beta site on update to master

[jglovier]

Expected behavior: When we merge pull requests to master, the system should deploy the latest to the beta site.

Actual behavior: We have to manually deploy to the beta site.

We can easily accomplish this with GitHub webhooks.

Some links for reference:

• https://developer.github.com/guides/automating-deployments-to-integrators/
• https://github.com/blog/2056-automating-code-deployment-with-github-and-azure
• https://github.com/blog/1241-deploying-at-github
• https://developer.github.com/v3/repos/deployments/

[ThomasWunderlich]

@jglovier Do you have a script you currently run to deploy? Or alternatively documented deployment steps? I went looking through the github repo and I wasn't sure what you're currently using for hosting or deployment of the beta site.

[jglovier]

@ThomasWunderlich it's pretty manual right now. I think @jkleinsc just logs into our host and git pulls basically LOL. 😄 cc @jkleinsc to confirm...

[ThomasWunderlich]

Well that at least confirms you're using a vm then and not docker. It makes the deployment process a bit easier

[jkleinsc]

@ThomasWunderlich @jglovier yup... right now I do the following:

1. Manual git pull on https://github.com/HospitalRun/hospitalrun-server
2. npm install
3. service hospitalrun restart

[jglovier]

Ideally we'd use something like Heaven so we can deploy via Hubot in Slack.

[parkr]

I would be interested in helping with this today. A v0.1 could replicate what @jkleinsc does above but via a cron job. That would be very simple to setup. Everything is public so it could be as simple as:

cat > /etc/cron.d/update_server <<EOS
[email protected]
0 * * * * root cd /data/hospitalrun-server && git pull origin master && npm install && service hospitalrun restart
EOS

A v1.0 could be:

• Stand up Heaven in Heroku. Give it a db & configure the apps needed to deploy.
• Configure webhook to hit Heaven. I believe it's just the Deployment event.
• Configure hubot-deploy in HospitalRun Hubot.
• ... anything else?

[parkr]

OK, after talking to @jglovier and @tangollama, it sounds like a general-purpose solution for deploying instances of hospitalrun-server would be better. For any instance that the HospitalRun team manages, it will be registered with heaven during the provisioning. Then, .deploy hospitalrun-server to prod would deploy via Heaven to all the hosted unites HospitalRun

• Setup a heaven Droplet. (Thanks @tangollama!)
• Install Docker (instructions used)
• Install Heaven via Docker (instructions used).
• Launch with docker run --name heaven --net heaven --publish 80:80 --env-file ./env.list -d emdentec/heaven
• Launch workers with docker run --name heaven-worker-1 --net heaven --env-file ./env.list emdentec/heaven "rake" "resque:work" "QUEUE=*" && docker run --name heaven-worker-2 --net heaven --env-file ./env.list emdentec/heaven "rake" "resque:work" "QUEUE=*"
• Configure Heaven to know about hospitalrun-server. This requires a deployment private key (it's SSH, right?) and perhaps a Capfile.
• Add hubot-deploy to HospitalRun's Hubot.
• Get hubot-deploy talking to Heaven.

[tangollama]

Note: @parkr is working on this.

[jglovier]

@parkr I started installing hubot-deploy on our existing [Slack hubot(https://github.com/HospitalRun/hospitalbot) when I realized you created the @hospitalrunbot user. Should I deprecate the hospitalbot repo in favor of that new bot user?

[tangollama]

@parkr @jglovier what do we need to do to finish this?

[tangollama]

@parkr @jglovier, consider this a very friendly bump. I'm happy to work some of this if that's feasible.

[jpeterson88]

Hey there, just bumping this to see where it currently stands @jglovier

[jglovier]

@jpeterson88 AFAIK there hasn't been any new progress on this since the last comments. If you're interested in addressing it, that'd be awesome.

@parkr anything @jpeterson88 should know from your previous work on this?

[jpeterson88]

Sweet. Yeah I'd love to take a stab at this. I see there is a discussion of using Heaven, are we set on that? Why not something like Kubernetes + Helm? We could hook into Hubot as well. Any additional info the team has would be much appreciated. Assuming I'll need credentials or whatever to where the site is hosted, who should I reach out to @jglovier ?

[MatthewDorner]

Sorry, clicked close button by mistake.

[jglovier]

I see there is a discussion of using Heaven, are we set on that? Why not something like Kubernetes + Helm?

I'm open to other solutions, I just proposed Heaven because it's what we used internally at GitHub and I was familiar with it a bit. But I'm not actively working on this issue and not strongly opinionated as to implementation details, so definitely open to other solutions that make sense. 👍

Assuming I'll need credentials or whatever to where the site is hosted, who should I reach out to @jglovier ?

Good question. @jkleinsc or @tangollama would have credentials for those.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[fox1t]

@tehkapa it's your duty!

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.