I found the error 
I was setting up some Mem and CPU limits and requests for the Gitlab Runner, unfortunately, the Cache step needed more CPU & Mem than the limit and the container was killed by Kubernetes without any log.
If anyone has the same problem… Check your requests & limits