Nikita Kazmin

Here's my quick-and-dirty benchmark with results: ``` ini BenchmarkDotNet=v0.13.1, OS=Windows 10.0.19044.1706 (21H2) AMD Ryzen 5 5600X, 1 CPU, 12 logical and 6 physical cores .NET SDK=6.0.300 [Host] : .NET 6.0.5...

Ok, so far it looks like this issue is more related to runtime+windows, and not Npgsql. I think the best way forward is to try to reproduce the same issue...

Yep, was able to reproduce. Will take a look...

I'm not sure I follow: whenever a server isn't available, it's supposed to be marked as `ClusterState.Offline`, and in that case it's going to be skipped until the cluster's state...

> @vonzshik looking at the logic in https://github.com/npgsql/npgsql/pull/4190/files#diff-d038284bfe7bf5480af7622a5fab46886b058fa3834d378d4b50e2770bf1b682R242, won't this always avoid the first host after a host that's down (except if it happens to be the index after 20...

> If both of this cases are reported as NpgsqlException with inner exception of type TimeoutException it is impossible to distinguished them. ~~That's definitely not how it works. For query...

@roji everything's green on my end...

@MichalUssuri I've been running you example for a while, and it doesn't look I can repro it. Just to make sure it's not PG taking this long to process the...

Hello. That's weird, since we do have a test to validate this behavior https://github.com/npgsql/npgsql/blob/86cd490d90092339c0ef448c020ee6ae1c0eefd3/test/Npgsql.Tests/CommandTests.cs#L161-L173 It does call `ExecuteNonQueryAsync`, but changing it to `ExecuteReaderAsync` still throws an exception. I've made a...

@adityamandaleeka here's a quick implementation I've made (it probably can be made faster, but just as an example): ``` ini BenchmarkDotNet=v0.13.2, OS=Windows 10 (10.0.19044.2130/21H2/November2021Update) AMD Ryzen 9 5900X, 1 CPU,...