"SQLSTATE[HY000] [2006] MySQL server has gone away" error

[Sébastien Mély]

Hi,

My SQL Cloud Instance had a problem for ~20 hours (13:27 GMT yesterday to 09:47 today), it randomly dropped connections with the following error, without any reason :

SQLSTATE[HY000] [2006] MySQL server has gone away

The graphs were also dead for this period :

I've found these errors in the logs, occurring only at the beginning and the end of this abnormal situation :

E 2018-12-18T13:27:08.099086Z 583928 [Note] Aborted connection 583928 to db: 'unconnected' user: 'root' host: 'localhost' (Got an error reading communication packets)
E 2018-12-18T13:27:09.334641Z 583904 [Note] Aborted connection 583904 to db: 'unconnected' user: 'root' host: 'localhost' (Got an error reading communication packets)
E 2018-12-18T13:27:09.335742Z 24 [Note] Aborted connection 24 to db: 'unconnected' user: 'root' host: 'localhost' (Got an error reading communication packets)
E 2018-12-18T13:27:11.526683Z 9 [Note] Aborted connection 9 to db: 'unconnected' user: 'root' host: 'localhost' (Got an error reading communication packets)
E 2018-12-19T09:47:58.863151Z 594976 [Note] Aborted connection 594976 to db: 'unconnected' user: 'root' host: 'localhost' (Got an error reading communication packets)
E 2018-12-19T09:47:58.864093Z 594975 [Note] Aborted connection 594975 to db: 'unconnected' user: 'root' host: 'localhost' (Got an error reading communication packets)
E 2018-12-19T09:47:58.914624Z 594974 [Note] Aborted connection 594974 to db: 'unconnected' user: 'root' host: 'localhost' (Got an error reading communication packets)
E 2018-12-19T09:48:06.935322Z 594973 [Note] Aborted connection 594973 to db: 'unconnected' user: 'root' host: 'localhost' (Got an error reading communication packets)

My project only uses the cloudsql/gce-proxy docker image to connect to the sql server.

Any idea what could have caused this situation ?

[amit...@google.com]

Hello Sébastien,

From you given timeframe and logs, it seems like your are affected by a known issue where Google Cloud Networking in the zone ‘europe-west1-b’ experienced a disruption. If your instance reside in this zone, most likely the issue was caused by the incident reported here. However without investigating your project, it is hard to state any conclusive answer. So if you believe this may not be the reason, I would recommend to create a private issue in issue tracker including your project ID and details of the incident so we can investigate further.