Gitea Actions cache server not configured (actions/cache fails) in CI

By Daniel Zoghalchali·Latchkey

actions/cache on Gitea talks to the act_runner built-in cache server. If the cache server is disabled or its advertised address is not reachable from the job container, cache steps error or silently do nothing.

What this error means

A cache step fails to reach the cache service, times out, or reports the cache API is unavailable, while the rest of the job runs.

Failed to restore: getCacheEntry failed: connect ECONNREFUSED
# act_runner cache server not enabled or not reachable from the container

Common causes

The runner cache server is disabled

act_runner ships a cache server that must be enabled in its config; when off, actions/cache has no backend.

The advertised cache address is unreachable

If the cache host/port the runner advertises is not routable from inside the job container, cache calls fail to connect.

How to fix it

Enable and address the cache server in config

1. Enable the cache server in the act_runner config.
2. Set an external address the job container can reach.
3. Restart the runner and re-run the cache step.

cache:
  enabled: true
  host: ""            # bind address
  port: 0             # auto
  external_server: "" # reachable URL if containers cannot reach host

Make the cache reachable from containers

When jobs run in docker mode, point the cache external address at a host/IP the container network can reach.

cache:
  enabled: true
  external_server: "http://172.17.0.1:8088"

How to prevent it

• Enable the cache server as part of runner config.
• Verify the cache address is reachable from job containers.
• Test a cache save/restore after changing runner networking.

Frequently asked questions

Related guides

References

• GitHub Actions documentation
• Workflow syntax for GitHub Actions