Postgres connection state stays active when an exception occurs

Issue

The connection state in pool is active even after the close/dispose methods are called. This issue is replicated only when you get an exception while executing a query or stored procedure.

Run the following code:

static void Main(string[] args) {
	NpgsqlCommand cmd = null;
	NpgsqlDataAdapter sda = null;
	NpgsqlConnection conn = null;
	try {
		string sql = "a_test";
		conn = new NpgsqlConnection("Server=localhost;Port=5432;Username=admin;Password=test;Database=majordb;SearchPath=dbs;CommandTimeout=60;MaxPoolSize=500;Connection Idle Lifetime=180;Connection Pruning Interval=5;");
		cmd = new NpgsqlCommand(sql);
		cmd.CommandType = CommandType.StoredProcedure;
		sda = new NpgsqlDataAdapter(cmd);
		cmd.Connection = conn;
		sda.Fill(dataTable);
	} 
	catch (Exception e) {

	}
	finally {
		if(null != sda)
		{
			try
			{
				sda.Dispose();                        
			}
			catch (Exception)
			{
			}
		}
		try
		{
			cmd.Connection.Close();
			cmd.Connection.Dispose();
		}
		catch (Exception)
		{
		}
		try
		{
			cmd.Dispose();
		}
		catch (Exception)
		{
		}
	}
}

To replicate the exception, I have create a test procedure that has pgsleep() greater than commandtimeout.

The Details

The issue details are also listed here in the below link: https://stackoverflow.com/questions/65629984/postgres-connection-state-is-active-when-an-exception-occurs

If the above code executes properly without any exception, the connection state in pool goes to idle, which is correct. But if an exception occurs while executing, like below:

"Npgsql.NpgsqlException (0x80004005): Exception while reading from stream ---> 
System.IO.IOException: Unable to read data from the transport connection: A connection attempt 
failed because the connected party did not properly respond after a period of time, or established 
connection failed because connected host has failed to respond."

The connection state in pool shows as active for about 5 mins or so, even though the close/dispose methods are called in finally block. This means the close/dispose did not properly executed by Npgsql. If the program keeps the connection state in pool active for every connection ran within 5 mins, then there can be an issue with MaxPoolSize error.

I wanted to see the connection state to idle, even when there is an exception. How do I do this.

Please note: I am not looking for a solution to the exception that I listed above. I am looking for a solution where the connection state is changed to idle from active when there is an exception while executing the above code.

To know if the connection state is active or not I used the following query:

SELECT
    pid,
    usename,
    application_name,
    datname,
    client_addr,
    rank() over (partition by client_addr order by backend_start ASC) as rank,
    state,
    state_change,
    current_timestamp,
    query,
    query_start,
    backend_start,
    FROM 
    pg_stat_activity
WHERE
    pid <> pg_backend_pid( )
AND
    application_name !~ '(?:psql)|(?:pgAdmin.+)'
AND
    datname = current_database() 
AND
    usename = current_user 

I have took screenshot for your reference for the above query.

Further technical details

Npgsql version: 4.1.7 PostgreSQL version: “PostgreSQL 11.8 on x86_64-pc-linux-gnu, compiled by gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-11), 64-bit”

[vonzshik]

It’s not related to #3446. The problem here is, the query executes for far too long, we fail with a timeout and the connection is broken. The query still executes on the backend, as the cancellation was implemented with 5.0.

[vonzshik]

@roji it’s 4.1.7 - there is no cancellation on the backend on timeouts until 5.0.