GitHub Actions "The runner has received a shutdown signal" / Lost Connection
By Daniel Zoghalchali·Latchkey
A job died because the runner lost its connection to GitHub partway through - the host crashed, was OOM-killed, a spot instance was reclaimed, or the network dropped. The job is marked canceled or failed without a clean error.
What this error means
A running job stops abruptly with "The operation was canceled" or "lost communication with the server", often after a period of no output. Re-running frequently succeeds, which points to an infrastructure blip.
Actions log
The runner has received a shutdown signal.
The operation was canceled.
# or
Error: The self-hosted runner lost communication with the server.
Common causes
Host crashed, OOM-killed, or spot-reclaimed
If the runner VM runs out of memory, panics, or is a spot/preemptible instance that gets reclaimed, the agent dies mid-job and GitHub reports lost communication.
Transient network drop
A brief loss of outbound connectivity breaks the runner long-poll. GitHub cancels the job once the heartbeat is gone for too long.
How to fix it
Add retries and right-size the host
Re-run failed jobs; transient drops usually pass on retry.
Give memory-heavy jobs a larger runner so they are not OOM-killed.
For spot/preemptible runners, handle reclaim with an autoscaler that re-queues the job.
Harden the runner connection
Ensure stable outbound HTTPS and keep the runner agent updated so it reconnects cleanly after blips.
How to prevent it
Right-size runners so memory-heavy jobs do not OOM the host.
Use stable on-demand runners for critical jobs, or auto-retry on spot reclaim.
Monitor runner host health (memory, network) alongside job logs.
Frequently asked questions
What causes "Runner lost connection"?
If the runner VM runs out of memory, panics, or is a spot/preemptible instance that gets reclaimed, the agent dies mid-job and GitHub reports lost communication.
How do I fix Runner lost connection?
Add retries and right-size the host
Can Latchkey fix this automatically?
Yes. Latchkey runs your GitHub Actions on managed runners that detect this failure, apply the fix, and retry the job automatically - self-healing is on by default.
