The Key to Go Efficiency is Just a Few Go Runtime Metrics Away! - Bartłomiej Płotka & Arianna Vespri

Go runtime metrics provide crucial insights into application behavior, memory usage, and garbage collection that can’t be obtained through regular profiling alone

Key default metrics to monitor:

• GoGC - controls garbage collection frequency and memory overhead
• GoMaxProcs - number of OS threads for Go routines
• GoMemLimit - maximum memory limit for the runtime
• GoMemStubHeapObjects - count of allocated heap objects
• GoRoutines - number of active goroutines

Setting appropriate GoMemLimit helps prevent OOM kills in containerized environments and allows better memory management

Runtime metrics help detect:

• Memory leaks
• Goroutine leaks
• GC pressure
• Stack overflow issues
• Scheduling latency problems

The Prometheus client Go library provides standardized collection of runtime metrics with filtering options to select specific metrics

Renaming metrics should be avoided as it breaks existing alerts, dashboards and automation

Both developers and SREs should monitor runtime metrics - it’s a shared responsibility for application reliability

Runtime metrics are essential for:

• Capacity planning
• Resource optimization
• Cost reduction
• Performance troubleshooting
• Production incident investigation

Default runtime metrics provide a good balance between observability and overhead - collect additional metrics selectively based on needs

Consider workload characteristics (CPU vs memory bound, concurrency level) when tuning runtime parameters like GC frequency