docker host with 1GB RAM: my descent into madness and cgroups

Guys I thoght this would be easy. Recieved a Time4VPS 1GB special, wanted to run my standard stack: cPanel + WordPress + mysql + redis + node app.

First week: OOM kills every 2 hours. Added swap. Swap thrashing made load avg 40.

Second week: Limited mysql to 256M cgroup. Mysql died under any traffic. Raised to 384M. Node app died instead.

Third week: Replaced mysql with sqlite. Redis with file cache. Node with static build. Still cgroups fighting over scraps.

Fourth week: I deleted everything. One static binary, 12MB, serves from RAM with 6MB free. Load avg 0.02.

Margins on this box? Negative if I count hours. Any leads on panel that runs in 50MB?

This thread is my tombstone

Actually this is the classic actually Resource starvation cascading failure pattern from the actually linuxKernelDocumentation or? When you actually configure the ControlGroups without actually understanding the Memory hierarchy you actually create a actually WorseThanNoLimitsSituation because the actually OOMKiller behavior becomes actually unpredictable and actually the kernel actually spends more actually cycles managing limits actually than running workloads actually or?

I have seen this actually in Production environments actually many times actually.