Connection leak happening in supavisor connections

[LakhveerChahal]

Bug report

• [N] I confirm this is a bug with Supabase, not with my own application.
  I'm not aware of the exact RCA, hence can't tell. I've tried SingleConnectionDataSource & PGSimpleDataSource in JDBC, but the issue still persists.
• [Y] I confirm I have searched the Docs, GitHub Discussions, and Discord.

Describe the bug

When connecting to Supavisor in Transaction pooler mode via JDBC (SingleConnectionDataSource or PGSimpleDataSource), the supavisor connections are not being released after calling connection.close(). I kept the suppressClose to false to avoid any connection proxy. After a while, the supavisor connections in grafana reaches to 200 count and application starts throwing below exception:
FATAL: Max client connections reached

On top of this, even if the application server is stopped, the supavisor connection count still remains the same. So, even if there was possibility of having bug in my application code, the connection count should ideally drop to zero when the server is killed.

Expected behavior

Supavisor connection should get released when connection is closed by application code.

Screenshots

System information

• OS: Linux (Although not applicable)
• Browser (if applies) Not Applicable
• Version of supabase-js: hosted on supabase
• Version of Node.js: Not being used
• Version of Java: 17

Additional context

References:
SingleConnectionDataSource - https://docs.spring.io/spring-framework/docs/current/javadoc-api/org/springframework/jdbc/datasource/SingleConnectionDataSource.html
PGSimpleDataSource- https://jdbc.postgresql.org/documentation/publicapi/org/postgresql/ds/PGSimpleDataSource.html

[linear]

POOLER-138 Connection leak happening in supavisor connections

[hauleth]

Is it self hosted Supavisor instance? If so, which version of Supavisor do you run?

[LakhveerChahal]

No, it is not self hosted. Using the Supabase hosted Supavisor only.

[hauleth]

@LakhveerChahal have you reported that to the Supabase Helpdesk? That would help us as it would provide us the information about your project so we would be able to access the logs and metrics for your organisation.

[jayeshgk95]

Hi @hauleth
We have added a support ticket with the following ID: 20158080681

[jayeshgk95]

@hauleth Do we have any update on this issue?
We are facing this issue on Project Id: xylvebryvatsekgukvlr

Adding a Screenshot of the "Number of Client Connections VS Time" graph for your reference.

[hauleth]

I have restarted your pool, so it should be now possible to reconnect. I will investigate it further, but I will move conversation to the support ticket.

[hauleth]

Ok, we have noticed the problem and I am working on a resolution. I am reopening. cc @chasers