> > > > The ASF does not run Travis. We pay a large amount of money TO Travis for > additional workers on top of the free tier. The reason I ask for discussion > regarding Jenkins usability is because throwing infinite money at solutions > like Travis is not an option. Now you find yourself in the same position > with GA, in that the free tier is insufficient. This is why I would like to > focus on improving ASF provided resources before considering throwing > infinite funding at vendors when there are serious known security problems, > lack of communication, and lack of insight into how the resources are used. >
To be perfectly frank - I would never ask for that. Throwing "ASF" money for a service that each of the "projects" can use as they want (and block others) without any control and monitoring is a recipe for disaster. This is not at all anything I would suggest. It does not solve the problem it just defers it in time at a huge cost. Bad idea. I think each project should have equally a 'fair' amount of resources AND a capability of adding their own resources if they need - but then they have to manage and optimise them on their own. There is no other way to make the money used well. We already plan our own optimisations once we have self-hosted runners in place. I have a great deal of experience in optimising CI. For one of my clients I managed to build an infrastructure, where for every single PR we built from scratch full Android OS (Lollipop) with 470 repositories in < 20 minutes (original time > 2h) and we cut down the cost from 5000 USD/month to 500 USD/month. We've done that by implementing auto-scaling, ephemeral AWS cloud instances with huge memory and building everything in-memory in tmpfs. CI has GREAT capabilities for optimisations, but it takes time and effort, and if you have infinite resources, you will have no incentive to do the optimisations even if it is possible. This was only possible because we monitored our own resources and we knew we had to spend our money well. I think from AWS putting pressure on GitHub to improve is one thing that can be done. And we do not really need them to cover the cost of the infra. We can get our own funds. No problem with that! With having 100s of projects, publicity, ASF brand, all those projects that can promote, and endorse GitHub and GitHub Actions is something that can be leveraged. And this is what I am asking for. We get from Github Actions is fantastic integration with GitHub which all our development effort focuses around. This is priceless when it comes to engineering time, building community and streamlining development efforts. If you have bad CI bad things happen in community building and cooperation - I started to snap recently at my collegue who started to complain that my PR is not ready yet, where the sole reason was slow CI. This is bad. So if only we had 'approved', "secure" and easy way of running our own self-hosted runners + a way from Github to distribute the free resources in a fair way among the project. - the problem would be immediately solved. This is really what I am asking for. > > > -Chris > > -- +48 660 796 129
