i've a question. We were using Gitlab's Instance runners but for security reasons, we wanna move it to private runner. I tried using ec2 as runner but it can't scale and is so slow for concurrent runs. I'm thinking to try GitLab Runner Autoscaling but do you have any other suggestions/idea? All responses are appreciated .
Best practices
1 Reply
The cost you pay in the cloud when building with scaling gitlab runners is:
- the cost to spin up new EC2 vms
- the cost to fetch the git repository again
- the cost to fetch build artifacts again
- for docker builds, the cost to fetch images again
- for docker builds, the cost to rebuild cachable layers
4 and 5 are solved by having a persistent docker build server. 3 is solved by having a persistent artifact cache 1 and 2 are solved by reusing the same gitlab runner instance.
There is no substitute, if performance is your goal, to hosting a large server that is capable of hosting all your builds in parallel. this gives you the immediate docker and gitlab artifact caching that is necessary to minimise redundant builds, maximise use of built artifacts and fetch network artifacts the minimum number of times.
that is all. on prem. or in the cloud. but expect to pay in the cloud.