Do you have the API call you used to create the repository? It seems that Elasticsearch itself cannot complete the request.
If you dial up Curator's logging to DEBUG level, you will see the full API call made to start the snapshot. If you were to execute that API call using curl or some other direct REST method, you would get the same result. It appears that Curator is not the problem if you're getting INTERNAL_SERVER_ERROR type messages (which are coming directly from Elasticsearch).
Either the shards are missing/unavailable, or the node in question ("node_id":"GOuzdEwtQ66KIuBpzqbXNg") is unable to contact/reach the repository, or something worse, perhaps. I can't tell from just these messages, but that's just not a good sign.