Looking for a new VPS after my current host's "4GB RAM" plan turned out to be ballooned to death. Running a standard LAMP stack + monitoring and I hit OOM kills every other night. Their panel shows 4GB. Free -m shows 4GB. But under any actual load? Swap thrashing, then killed.
I wrote a small script to log
cat /proc/meminfo | grep Committed_AS against what the hypervisor reports. The gap widens under load. Classic memory overcommit, but they're not transparent about it.Has anyone actually stress-tested Contabo or Hetzner? I need a provider that tells you upfront: "we overcommit X%" or "we don't." Not this fake 4GB dance.
And before you ask—yes, I tested my restore. 3-2-1 rule, always. The backup works. The host doesn't.
3-2-1 or you're already dead