Slack notifications from Github Actions

A while back I wrote about moving my deployment scripts away from Codeship into Github Actions. This has continued to work, but as I noted it did leave me without notifications in Slack. Time to fix this.

There are a few ways of enabling notifications. The best way involves creating Slack apps with incoming webhooks, but this involves a lot of faff (especially on a free Slack instance). The ever-useful Phil Wilson found a much simpler alternative.

Enter the Github Slack app.

The setup is very easy:

• Add Github app to Slack
• Log in to Github app as whichever user should be receiving notifications
• Authorise it to talk to your Github account in Github

All three of these are documented in the above link.

Then it's a case of setting up the notifications you want within the Slack app channel. I used the following:

/github unsubscribe $account/$repo issues pulls commits
/github subscribe $account/$repo workflows:{event:"push"}

The first removed the overly noisy notifications from a whole bunch of Things on the account. The second enables notifications from workflows (ie Actions) - mine are triggered on "push" events. Docs for the workflow configuration can be found on the integration page.

And behold! Notifications!

Migrating away from Heroku

Along with almost every developer I know, I was very sad to see the end of Heroku's free tier. I was hoping they'd reverse this decision, but it seems that was too optimistic so in the end I had to choose between killing all my toy applications, paying money or finding an alternative.

I did consider just shutting things down, but I had two concerns with this. First, I have one application I actually use and losing it would be a pain. Second, and more importantly, if I have no solution to this problem it adds another barrier to me coding as a hobby. So this needs solving, and not just by throwing hundreds of pounds a year at it.

Finding an alternative took a long while, primarily because I was trying to find a straight all in one alternative. While they are out there, the free tiers are not suitable - the closest insist on deleting my database after 90 days. Instead, I ended up separating application hosting, database hosting and sending of email and the below is where I ended up, written in detail in case it helps one of the many others doing this same thing. I did this all back in November, before the Heroku switch-off, and a few months all it seems to all be running well.

My applications

Everything I'm moving is a personal project (no url customisation, limited need for backups, high availability, etc) and a simple 12f(ish) application. They are written in Ruby (mostly Rails, one Sinatra) and make use of Postgres on the backend and some simple mailing. Nothing complicated and Heroku managed all this for me before.

To move from Heroku SendGrid to independent SendGrid, I did find I had to make a small update to the mailing config.

Most of this diff are linting updates - only the domain change should be necessary.

Databases

I chose ElephantSQL as it has a good enough free tier offering (limited to 20mb) and lets the free databases persist. Detailed steps for anyone who, like me, hasn't done much with a Postgres database for a while.

Setup:

• Create account in ElephantSQL (sign in with Github)
• Create database
• Get connection string

Backup from Heroku:

heroku pg:backups capture --app myapp
curl -o latest.dump `heroku pg:backups public-url --app myapp`

Restore to ElephantSQL:

pg_restore --host "machine.db.elephantsql.com" --port "5432" --username "blah" 
    --password --no-owner --dbname "blah" --clean --verbose "latest.dump"

And voila, database ready to rock in ElephantSQL. This can be checked by running up locally, thus:

DATABASE_URL=postgres://blah:blahblah@machine.db.elephantsql.com/blah bundle exec rails s

Mailing

My requirements are very light here since I only use email for "forgotten passwords" email. For this, a free [SendGrid](https://sendgrid.com/) account was more than enough. I'm using SendGrid because this was the Heroku solution, so it makes for an easy switch.

• Create account in SendGrid
• Create API key

Application hosting

I wanted something which, like Heroku, was a container-based PaaS. I certainly can wrap my applications in containers or deploy to virtual machines, but honestly who has the time. I want to point my application at a hosting provider and have it do the work for me.

For this, I looked at a load of options but I settled on two - Koyeb and Render. Both have workable free tier hosting. Render works like Heroku used to - giving a number of hours per month and spinning down applications when they are not in use. Koyeb simply gives you $5 credit per month and lets you choose how to host. This is an enough to have an application running on a low tier package without spinning down, but not enough to run several applications.

For deployment, the steps were basically the same. For Koyeb:

• Create account in Koyeb (sign in with Github)
• Create application
• Add Github integration
• Add to "run" command: rails db:migrate && rails server
• Select repo (this took a while to appear for me, although Github was having issues when I did this)
• Add DATABASE_URL env var
• Add RAILS_MASTER_KEY env var
• Add SendGrid env vars
• And any other env vars

Then off it goes. I did find it took a little while to start working, but it has been fine since. Also, a slight gotcha - Koyeb appears to have some broken validation when it comes to counts, eg "must be 3 characters or more" seems to actually mean "more than three characters".

For testing, I did all this (including the mailing change above) in a branch then cleaned up at the end.

For Render it was much the same, except I didn't need to specify the RAIL_MASTER_KEY env var and I needed to write a build script.

Cleanup

Finally, to finish off:

• Spin down the Heroku app
• Remove auto-deploy to Heroku (in my case from Codeship)
• Toggle Koyeb / Render over to watch "main" if using a test branch
• Make sure the instance is healthy (and make sure not exposed the wrong port anywhere)
• Final redeploy

Roundup

That is pretty much it. I have been running this setup since November with no problems. I am getting fewer notifications through to Slack, so at some point I would like to get better alerting of start / end deployments but since it's just me working on these things I can watch them easily enough - it's not a priority. I'm just pleased I can still deploy easily and free as that means I'll keep my hand in writing bits and pieces.

Buying code

TL;DR - I found this old post about outsourcing and found reading it again interesting.

EDIT - for reasons I choose not to speculate on, the original post has been archived and must be seen via the Wayback machine.

The magic bullet that is outsourcing rears its head every so often. Why worry about organisational capability and capacity when you can commoditize the process? Push some money at the problem and it goes away, right?

Well, no. Not even close. There are many problems fundamental with this way of thinking and for me they boil down to two key things:

1. Code is never finished
2. Effective and sustainable outsourcing is really hard

I'm going to write a separate post about the first point. Probably more than one - every time I sit down to write something it grows by a frightening amount. The short version is that written code is the start of the journey, not the end. It's why we have entire frameworks (such as ITIL) to manage the services created. You may have paid someone to create the software - but then what?

I was going to write a series of thoughts to cover the second point. But then I discovered something fun - I've been writing on the internet long enough to be repeating myself. So here is me, nearly six years ago, writing about the pitfalls of outsourcing on my old work blog. It's worth a click - if only so they see a weird spike in traffic on an older post.

I find the thoughts there interesting for a few reasons. Firstly, the world hasn't really changed. At all. The fact that I'm writing this post is due to this approach still being seen as an easy solution, and still having some very dangerous long-term effects. The only point in there I think has improved is that Agile working with a supplier is a bit easier these days - although that might be a reflection of the difference in agencies operating in London and Bath.

Secondly, I find it interesting that I felt strongly enough back then to write a post about it. I was a developer, writing code and helping shape a university department. I saw problems with outsourcing, raised them, and when things happened anyway (and, despite the glib phrasing, I should make it clear that my concerns were addressed in the context of the department's priorities) I was one of the people who was saddled with the fixing and learning to make everything work.

These days I'm the Head of Software Engineering for GDS (for the moment at least) and I get to inflict that fun on other people. Or, more accurately, I get significant input into the operating model for the department - which means I need to make sure we're taking into account the points past-me made and ensure we're effectively taking them into account when considering how we build things. Because otherwise, I'm inflicting that fun on other people.