CircleCI "Too long with no output (exceeded 10m0s)" - Fix

By Kaveh Alemi·Latchkey

CircleCI cancels a step that goes 10 minutes without writing anything to stdout/stderr. It is a safety timeout against hung jobs - the step may still be working, just silently.

What this error means

A long step (a big download, a slow test, a quiet build) is cut off with "Too long with no output (exceeded 10m0s): context canceled", even though it was making progress. Re-running sometimes passes if timing shifts.

Too long with no output (exceeded 10m0s): context canceled

Common causes

A genuinely slow but silent step

A large dependency install, dataset download, or compile runs longer than 10 minutes without printing, so CircleCI’s default no_output_timeout cancels it.

Buffered output that never flushes

A tool buffers its logs and emits nothing until the end. CircleCI sees no output and assumes the step hung.

A real hang or deadlock

The step is actually stuck - waiting on a lock, a prompt, or a network call that never returns - and the timeout correctly fires.

How to fix it

Raise no_output_timeout for legitimately slow steps

- run:
    name: Integration tests
    command: ./run-slow-tests.sh
    no_output_timeout: 30m

Emit periodic progress

Make the step write something regularly so CircleCI sees it is alive.

- run: |
    long-task &
    pid=$!
    while kill -0 $pid 2>/dev/null; do echo "still running..."; sleep 60; done

Unbuffer or add verbose flags

1. Enable the tool’s progress/verbose output so it prints periodically.
2. Wrap chatty-but-buffered commands with stdbuf -oL / unbuffer.
3. If it is a true hang, fix the underlying lock or network wait.

How to prevent it

• Set a realistic no_output_timeout on known-slow steps.
• Keep long steps emitting progress so the timeout never trips.
• Investigate repeated timeouts as possible deadlocks, not just slowness.

Frequently asked questions

Related guides

References

• GitHub Actions documentation