Basically my problem is that a few plugins working together exhausted the file table.
We are developing a medium sized API with ~90 resolvers, where ~50 of those are lambdas. We hit the AWS stack limit quite fast. So we head-on to the next plugin.
This solves our resource limits problem but introduces a parameter limit one, requires an unreleased feature branch to make it work. No big deal.
We go on developing with a single bundle, and hits the 50MB limitation in no time. Next plugin!
This is the most helpful, reduces about 30% of the size in final bundle. But we still hit the 50MB hard within a week. This time we chose not to go with plugins, but instead a built-in feature.
This splits the monolithic bundle into 1.5MB packages, solves most of the problems while brings the most headache to the team.
Our GitLab runner merely builds our 50-ish lambda, barely optimized it. But then the excluding development at once always overflows the file table.
We don’t mind waiting 35 mins of deployment each time in production, but at least it needs to be done.
Is there some way to restrict the lambda bundle process into, say, 10 processes each time, instead of flooding all of them in parallel?