We were trying to move our Selenium tests into a docker container and were getting the above error response. First Google suggested increasing the memory of the container to 2G. We increased it to 3G to no effect. Deeper Googling then suggested increasing shared memory. Initially, it was 64m. We raised it to 256m and it magically worked! Our script for creating the container:
So this is all about getting started with AWS. I’ve been using AWS tangentially for about five years, almost always EC2 instances, so not really pushing the envelope. I’d really like to get my head round serverless and lambdas but I’m having a bit of a conceptual problem trying to work out a use case to do at home.
I have done the Udemy “AWS Certified Cloud Practitioner Practice Exam” which was quite frankly brutal, demanding 90% for three papers. First time through each I got 70-80% which I thought wasn’t too shabby but obviously not good enough for their arbitrary cutoff. The second time through I got >90% which was nice.
The biggest takeaway I have is once you’ve created a root account, create a user account and only give it only the privileges it needs. Security you know!
This is the list of white papers I’ve ingested to far. I hope it proves useful.
This is the motherlode. If you want an overview of all available services on AWS, this is the place to start. For us in the perl world, that’ll be git, CI/CD pipelines and EC2/Fargate. If you’re jiggy, docker too. At this point for most of that, I like Gitlab, especially since M$ took over github.
AWS Well Architected Framework
This takes the overview one step further. Apparently as you partition your app vertically, VPSs are the trick. Security again.
In a good chunk of the contracts I’ve done, the developers have been good boys and written tests, but there was no way of automatically running them. Jenkins fixes this. Better still using the Perl TAP output formatter we can get a nice graph of the number of tests increasing. If you like, you can run Bamboo or GoCD but I’ve not had happy experiences with either of these.
Practising Continuous Integration and Continuous Delivery on AWS
Taking Jenkins one step further. In an ideal world code gets committed to master and then gets made live. Your process may vary. Oh, and database versioning is Hard. I’m reliably informed squitch is the one for this.
Now we start getting to the docker meat. I’m not sure how applicable this is to a clunky monolithic Perl framework. I dockerised a simple Catalyst app and it was HUGE. Back to CGI.pm?
A common whinge I’ve heard is that unless you’re careful and out of the free first year tier, is that suddenly your AWS usage blooms into thousand of pounds a month. Having been charged £15 a month for static IP I can well believe it.
Kubernetes land is still a mystery to me. Every way I’ve tried to approach it, from linux to Mac I’ve been thwarted. Oh well, one day it’ll be mature enough and actually work for me.
So that’s what I have so far. I think most of the time, we’ll go EC2 and then RDS. I’d put Cloudflare on the front unless I particularly needed a Route53 feature. Serverless is still in the land of dragons and Perl isn’t spectacularly well supported. I’d like to see a world where the code pipeline is under Amazon as well as horizontal scaling with the load balancers.
Recently I had a contract which required me to immerse myself in AWS configuration for a LAMP stack, which led me to Terraform, amongst other things. I thought I’d publish here the list of resources that helped me. It’s a lot of links!
One tip I found was to use Jenkins to do automatic validation of your Terraform scripts. I think that’s a good tip.
And there’s the link dump. It should be enough to keep you going for a couple of days! Overall, I think if you’re going to use AWS, embrace it fully. Except Cloudflare. They’re awesome.
