spawn EBADF on all exec calls in remote OpenClaw sandbox (macOS + Node 22) — local install works
Hi team @Krill 🦐 — I’m running into a consistent spawn EBADF failure, but only in the remote OpenClaw sandbox environment, not locally. I want to sanity-check whether this is a known sandbox/runtime issue or something actionable on my end.
Environment split (important)
I’m interacting with OpenClaw in two different environments:
Local (working):
Hardware: Mac mini (arm64)
OS: macOS 26.2
Node: 22.22.0
OpenClaw: 2.1.1 / 2026.1.30
Result: exec works normally (ls, printf, etc.)
Remote sandbox backing this chat (broken):
Environment: managed OpenClaw sandbox / container
Node / OS: unknown (platform-controlled)
Result: every exec call fails immediately with spawn EBADF
Symptoms
All exec calls fail instantly, even trivial ones:
printf 'hi'
ls
openclaw status
No useful stack trace beyond spawn EBADF
Failure happens before command logic executes
Restarting sessions does not help
What I’ve already tried
Restarted agent sessions (sessions stop/start) — no change
Reinstalled / updated OpenClaw locally — fixed local exec only
Verified same commands succeed locally but fail in sandbox
Avoided exec entirely in sandbox; reasoning and browser tools work fine
Key observation
This appears to be environment-level, not configuration-level:
Local OpenClaw is healthy
Remote sandbox cannot spawn subprocesses at all
Session restarts don’t rebuild the underlying runtime
Question
Is this a known issue with the current sandbox image (Node 22 + macOS/Linux kernel + sandboxed spawn)?
If so:
Is there a planned fix or sandbox recycle?
Is there a way to force a fresh sandbox allocation?
Or is exec intentionally unsupported in this environment for now?
Happy to provide logs or run diagnostics if there’s something specific you want tested.
Thanks — appreciate any guidance here.
Environment split (important)
I’m interacting with OpenClaw in two different environments:
Local (working):
Hardware: Mac mini (arm64)
OS: macOS 26.2
Node: 22.22.0
OpenClaw: 2.1.1 / 2026.1.30
Result: exec works normally (ls, printf, etc.)
Remote sandbox backing this chat (broken):
Environment: managed OpenClaw sandbox / container
Node / OS: unknown (platform-controlled)
Result: every exec call fails immediately with spawn EBADF
Symptoms
All exec calls fail instantly, even trivial ones:
printf 'hi'
ls
openclaw status
No useful stack trace beyond spawn EBADF
Failure happens before command logic executes
Restarting sessions does not help
What I’ve already tried
Restarted agent sessions (sessions stop/start) — no change
Reinstalled / updated OpenClaw locally — fixed local exec only
Verified same commands succeed locally but fail in sandbox
Avoided exec entirely in sandbox; reasoning and browser tools work fine
Key observation
This appears to be environment-level, not configuration-level:
Local OpenClaw is healthy
Remote sandbox cannot spawn subprocesses at all
Session restarts don’t rebuild the underlying runtime
Question
Is this a known issue with the current sandbox image (Node 22 + macOS/Linux kernel + sandboxed spawn)?
If so:
Is there a planned fix or sandbox recycle?
Is there a way to force a fresh sandbox allocation?
Or is exec intentionally unsupported in this environment for now?
Happy to provide logs or run diagnostics if there’s something specific you want tested.
Thanks — appreciate any guidance here.
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
83,815Members
Resources
Similar Threads
Was this page helpful?
