Google Cloud SQL Second generation -> "Aborted connection"

user_1931996 0
Viewed 1572

I am running a Java + JPA/Hibernate application on Appengine and switched my database from the first generation Google Cloud SQL instance to the second generation and now get a lot of this errors:

2017-05-20T22:49:53.533247Z 2235 [Note] Aborted connection 2235 to db:
'mydb' user: 'root' host: 'cloudsqlproxy~myip' 
(Got an error reading communication packets)

As far as I can tell, most of these error occur during database requests inside task queue tasks.

This did not happen with the first generation. How can this be avoided?

google-cloud-platform google-app-engine jpa hibernate sql
2 Answers

Τhe "Aborted connection nnnn to db:" message is triggered when an existing connection is terminated improperly as described by Google’s documentation. Most of the aborted connections happen because of the termination of your connection was not correct or because of networking problem between the server and the client, as described in the documentation here.

I advise you to follow the Google’s documentation about managing Cloud SQL connections, emphasizing on the “connection pools” section and of course the “opening and closing connection” section.

user_89634860

Managing database connections talks about "close connection properly". However, in my case, the error still occurs when I use GCP cloud function to connect GCP cloudSQL.

A Google group says that, unless you use NullPool OR dispose the engine explicitly, the error message will always occur. Also, it does not suggest to use engine.dispose().

So I wonder what is the best way to release the resource of connection pool without generating error message on CloudSQL?

user_116370160
Related