Doctor suggests setting NODE_COMPILE_CACHE and OPENCLAW_NO_RESPAWN for low power hosts.

[covert gale]

But what do those actually do? What is lost or gained by setting them? Version 2026.4.15.

[sleek kindle]

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

[dull eagle]

I checked the 2026.4.15 code, short version:

• NODE_COMPILE_CACHE=/var/tmp/openclaw-compile-cache

  • Tells Node where to keep its on-disk module compile cache.
  • Gain: repeated CLI runs get faster, especially status, health, doctor, --help, and other short commands on Pi/cheap VPS boxes.
  • Cost: a bit of disk usage and a few extra writes, and the first run mostly just warms the cache.
  • Subtle detail: OpenClaw already enables Node’s compile cache, so setting this mainly pins it to a better persistent location (/var/tmp instead of a temp dir that may get wiped).

• OPENCLAW_NO_RESPAWN=1

  • Disables OpenClaw’s extra self-respawn step.
  • Gain: slightly less startup/restart overhead, which helps on weak hosts.
  • Cost: you give up the automatic respawn pass that can inject some startup-only Node flags, mainly warning suppression and auto NODE_EXTRA_CA_CERTS setup. Gateway restarts also stay in-process instead of doing a fresh PID respawn.

So practically:

• Set both on a Pi / small ARM VM / low-power VPS, worth it.
• Ignore them on a normal machine, difference is usually minor.

• OPENCLAW_NO_RESPAWN=1 does not mean “don’t restart ever”, it just skips OpenClaw’s own self-respawn path. systemd/launchd restart policy is separate.

Docs:

• https://docs.openclaw.ai/vps
• https://docs.openclaw.ai/platforms/raspberry-pi

If you want, I can also give you the “safe default” recommendation for a 2026.4.15 low-power host in one copy-paste block.