Firebird ADO.NET-Provider connection pool error on connection lost

First, this is not a linq2DB error, it is a tip for a workaround caused by the current Firebird ADO.NET provider (Version 4.6.0.0 - 5.12.1.0) If you are working with connection pools and your connection is lost e.g. because the Firebird server is closed, the pooled connection is still marked as ‘opened’. So every database access will create an Exception (FBException) ‘Error reading data from the connection’. So if the Firebird server is accessible again you still cannot access it because the old connection will be used!

One solution is to disable pooling. I don’t like this because of performance reduction.

Another solution is to handle it in the OnTrace Action.:

DataConnection.OnTrace = delegate(TraceInfo info)
		{
			try
			{
				// First time access, if no connection has been ever established, this will throw an error!
				var conn = (FbConnection)info.DataConnection?.Connection;

				var fbException = info.Exception as FbException;
				// The error code stands for 'Error reading data from the connection'
				if (fbException != null && fbException.ErrorCode == 335544726)
				{
					// Clear all pooled connections, because the connection to the database is lost!
					FbConnection.ClearAllPools();
				}
			}
			catch (Exception ex)
			{
				// No database connection was ever opened!
				return;
			}
		};

[frankiDotNet]

This works:

	/// <summary>	A Firebird retry policy. </summary>
	public class FbRetryPolicy : RetryPolicyBase
	{

		/// <summary>	Not try again error codes, that clears the connection pool. </summary>
		public static List<int> NotTryAgainErrorCodes = new List<int>()
		{
			335544726, //	read_err
			335544721, //	network_error
			335544421, //	connect_rejec
		};
		
		/// <summary>	Default constructor. </summary>
		public FbRetryPolicy() : base(Configuration.RetryPolicy.DefaultMaxRetryCount, Configuration.RetryPolicy.DefaultMaxDelay) { }

		/// <summary>	Constructor. </summary>
		/// <param name="maxRetryCount">	Number of maximum retries. </param>
		/// <param name="maxRetryDelay">	The maximum retry delay. </param>
		public FbRetryPolicy(int maxRetryCount, TimeSpan maxRetryDelay) : base(maxRetryCount, maxRetryDelay) { }
		

		/// <summary>
		/// Determines whether the specified exception represents a transient failure that can be
		/// compensated by a retry.
		/// </summary>
		/// <param name="exception">	The exception object to be verified. </param>
		/// <returns>
		/// <c>true</c> if the specified exception is considered as transient, otherwise <c>false</c>.
		/// </returns>
		protected override bool ShouldRetryOn(Exception exception)
		{
			FbException fbException = exception as FbException;
			// The error code stands for 'Error reading data from the connection'
			if (fbException != null && NotTryAgainErrorCodes.Contains(fbException.ErrorCode))
			{
				// Clear all pooled connections, because the connection to the database is lost!
				FbConnection.ClearAllPools();
				return false;
			}
			return true;
		}
	}

In your context costructor just call:

this.RetryPolicy = new FbRetryPolicy();

[sdanyliv]

Never tried that, but there many unit tests. https://github.com/linq2db/linq2db/blob/cf160d92b7062d1af7581abeb7185b20a0fdde94/Tests/Linq/Data/RetryPolicyTest.cs