"ORA-50012: Pooled connection request timed out" when timeout transaction in >= .net8

[mishustin]

Hello!

I have a simple example that reproduces the error “ORA-50012: Pooled connection request timed out”. After the error occurs, the oracle connection pool does not issue any more connections and the application hangs without accessing the database, only restarting (i.e. resetting the oracle connection pool) helps.
The error is reproduced only on versions starting from .net8, on .net6 this error is not reproduced. It occurs if the transaction timeout is less than the code execution time inside the transaction.

OS: Microsoft Windows NT 10.0.19045.0
.net version: 8.0.12
configuration buils: release
Microsoft.EntityFrameworkCore Version 8.0.12
Oracle.EntityFrameworkCore Version 8.23.70
Oracle.ManagedDataAccess.Core" Version 23.7.0

Example
https://github.com/mishustin/OracleTransactionApp/tree/ORA-50012

Is this bug?
Can you help me?

[alexkeh]

I ran your test case. For the .NET 6 setup, I see the transaction timeout. However, the .NET 8 setup doesn't trigger a transaction timeout. I suppose that means the connection time out cannot occur.

Can you generate an ODP.NET level 7 trace file, which should also indicate why the connection time out occurs?

You can also try using the ODP.NET 21c version for EF Core 8 to see if that works. There have been several changes to connection pool behavior between 21c and 23ai, one of which could be the reason for the time out you observe.

[mishustin]

You should repeat the run a few times on .net8, usually on the second or third time there is a transaction timeout and then immediately an error getting a connection from the oracle connection pool.

In this version configuration, the error “ORA-50012: Pooled connection request timed out” does not really occur and connections are returned from the connection pool after other errors, that's okay

“Microsoft.EntityFrameworkCore” Version=“8.0.12”
“Oracle.EntityFrameworkCore” Version=“8.21.170”
“Oracle.ManagedDataAccess.Core” Version=“3.21.170”

Ironically, we switched to a new version of the library to add OpenTelemetry tracing to understand the cause of the “ORA-50012: Pooled connection request timed out” error

Trace is here

ORACLETRANSACTIONAPP.EXE_PID_17040_DATE_2025_02_14_TIME_13_38_54_136029.txt