SFTPClient hangs forever in Windows- and Linux-based Azure App Service

We use an ASP.NET Core 3.1 Background Service to watch and import files from an Sftp Server.

We encountered two issues:

• ListDirectoryAsync hangs forever (SSH.NET 2016.1.0)
• Connect hangs forever (SSH.NET 2020.0.0-beta1)
  • After running it for about 2 days successfully)

I created a little repro application were calls to ListDirectoryAsync and Connect are involved.

public class BackgroundWorker : BackgroundService
{
    private readonly ILogger<BackgroundWorker> _logger;

    public BackgroundWorker(ILogger<BackgroundWorker> logger)
    {
        _logger = logger;
    }

    protected override async Task ExecuteAsync(CancellationToken stoppingToken)
    {
        do
        {
            var host = "";
            var username = "";
            var password = "";
            var uri = new Uri(host);
            var path = "FI/Sample";
            var parsedHost = uri.Host;

            _logger.LogInformation($"{DateTime.Now} - Creating new client ...");
            var parsedPort = uri.Port;
            var connectionInfo = new PasswordConnectionInfo(parsedHost, parsedPort, username, password)
            {
                Timeout = TimeSpan.FromSeconds(120),
            };

            var client = new SftpClient(connectionInfo);

            _logger.LogInformation($"{DateTime.Now} - Client created.");

            client.OperationTimeout = TimeSpan.FromMinutes(3);
            _logger.LogInformation($"{DateTime.Now} - Connecting ...");
            client.Connect();
            _logger.LogInformation($"{DateTime.Now} - Connected");

            var stopwatch = new Stopwatch();
            stopwatch.Start();
            _logger.LogInformation($"{DateTime.Now} - ListDirectoryAsync ...");
            var immutableArray = (await client.ListDirectoryAsync(path))
                .Where(x => x.IsRegularFile)
                .Select(x => new { x.Name, x.LastWriteTimeUtc })
                .ToImmutableArray();

            _logger.LogInformation($"{DateTime.Now} - Done({immutableArray.Length}) {stopwatch.ElapsedMilliseconds}ms");

            _logger.LogInformation($"{DateTime.Now} - Disposing ...");
            client.Dispose();
            connectionInfo.Dispose();
            _logger.LogInformation($"{DateTime.Now} - Disposed.");
            await Task.Delay(1000);

        } while (true);
    }
}

Further findings:

• We could see that both Linux and Windows-based App Services hung at the same time. So it seems that the Sftp Server might have had an issue at that time. Nevertheless SSH.NET should handle this with a timeout.
• We could also see that if we increase our polling interval hanging process were less frequent.

If you need more information, let me know.

Update:
Seems like https://github.com/sshnet/SSH.NET/blob/cefdc203d98cd890815e029bc759bc43ec5a9643/src/Renci.SshNet/Session.cs#L1945 is the problem.

Issue に示されている再現用の BackgroundWorker から始め、リンク先の 1945 行付近にある Session.cs を調べ、Connect と ListDirectoryAsync のパスに焦点を当てます。Azure App Service のハングまたは利用できない SFTP サーバーを再現し、影響を受ける操作が無期限にハングせず、設定されたタイムアウトを遵守することを確認します。