Major Description
When a repository is set in mirrored mode, opening the usage logs sometimes errors out with a 500 and the following Python trace is observed:
{"log": "gunicorn-web stdout | json.decoder.JSONDecodeError: Unterminated string starting at: line 1 column 7671 (char 7670)\n", "stream": "stdout", "time": "2022-01-19T12:22:25.918649631Z"}
{"log": "gunicorn-web stdout | obj, end = self.scan_once(s, idx)\n", "stream": "stdout", "time": "2022-01-19T12:22:25.91864645Z"}
{"log": "gunicorn-web stdout | File \"/usr/lib64/python3.8/json/decoder.py\", line 353, in raw_decode\n", "stream": "stdout", "time": "2022-01-19T12:22:25.918643037Z"}
{"log": "gunicorn-web stdout | obj, end = self.raw_decode(s, idx=_w(s, 0).end())\n", "stream": "stdout", "time": "2022-01-19T12:22:25.918640311Z"}
{"log": "gunicorn-web stdout | File \"/usr/lib64/python3.8/json/decoder.py\", line 337, in decode\n", "stream": "stdout", "time": "2022-01-19T12:22:25.918638066Z"}
{"log": "gunicorn-web stdout | return _default_decoder.decode(s)\n", "stream": "stdout", "time": "2022-01-19T12:22:25.918635894Z"}
The issue stems from the fact that the metadata JSON stored in the logentry3 table is not correctly terminated so the decode JSON function fails at unmarshalling it.A sample of the errorneous logentry3 table row is added to the case in text format.
The only workaround that I see is to delete rows from the logentry3 table manually, connected to the mirror tag operation.
Please check!
