Skip to main content

Many Services

When service count grows, orchestration correctness is usually fine; developer feedback loop speed is the real constraint.

Typical pressure points​

  • long startup chains from dense dependsOn
  • readiness probes on too many non-critical services
  • high log throughput making incident analysis noisy
  • frequent full-stack restarts for single-service changes

Mitigations​

  1. Keep dependencies strict and minimal.
  2. Avoid serializing independent services into deep chains.
  3. Use targeted runtime commands (restart_service) instead of full restarts.
  4. Set sensible logView.maxEntries defaults for noisy services.
  5. Split optional subsystems into separate configs when they are not needed every day.

Signal that you outgrew one stack​

  • startup time becomes the main cost of each edit cycle
  • one team's actions frequently destabilize another team's flow
  • local machine thermal throttling is common during normal iteration

When these signals persist, split stacks intentionally instead of fighting one giant config.

See Waves and When to Split Stacks.