manage_vms needs to be quicker

google / clusterfuzz

Scalable fuzzing infrastructure.

https://google.github.io/clusterfuzz

Apache License 2.0

5.29k stars 554 forks source link

manage_vms needs to be quicker #4164

Open jonathanmetzman opened 2 months ago

jonathanmetzman commented 2 months ago

It's slowness is a major source of pain when updating oss-fuzz. We should give the instance running more cores and make more quota available for it.

oliverchang commented 2 months ago

Another potential bottleneck is rate limits from GCE.

Also, we should get rid of manage_vms completely once we migrate oss-fuzz to the new architecture (and Kubernetes).

vitorguidi commented 2 months ago

Another option is to bring it to terraform, and a kubernetes controller could take over the job of doing the oss fuzz core slicing and host assignment. Good opportunity to decouple business logic from infra management