NewWeekly newsletter with catalog picks and agent news · Read
ToolsMCPBlogResearchCommunityStar on GitHub
All Tools
A
Dev ToolsFreeOpen Source

AUTOMAXPROCS

Match GOMAXPROCS to container CPU quota

MIT

ABOUT

Go defaults GOMAXPROCS to the host CPU count, which is too high inside a cgroup with a small quota. Agent and inference pods then get throttled. automaxprocs reads the quota at startup and sets GOMAXPROCS to match so throughput and tail latency stay stable.

INSTALL
go get go.uber.org/automaxprocs

INTEGRATION GUIDE

1. Import as a blank import in a Go inference server running on Kubernetes 2. Cut CPU throttling on agent workers with a fractional CPU request 3. Align GOMAXPROCS with a 2-core quota instead of a 24-core node 4. Keep P99 latency down when serving LLM APIs from cgroup-limited pods

TAGS

gokubernetesgomaxprocscontainersuberperformance