Open captainsafia opened 2 years ago
Failed on Win11... Kestrel responded to the client-hello, but then the client never completed the handshake or sent the request. Related to our other Win11 issues?
Win11 was also the first thing I looked for. The super slow handshake does make me wonder if we're overwhelming the helix agents in the Win11 queue which might be contributing to the other failures. Of course, data corruption isn't acceptable even on an overburdened machine, but it might explain why the other issues have only appeared on this queue if it happens not to be platform-specific.
We still haven't seen data corruption in memory, so we can't rule out an issue at the socket layer. I'm going to look at updating some of the Interop.FunctionalTests to use the in-memory transport and ConnectCallback.
One big difference about this test, it's not using HttpClient or Sockets, only HttpClientSlim and SslStream in-memory. Or am I reading it wrong?
Ah, there is one difference on Win11 that will explain part of this: TLS 1.3 is enabled by default, and it uses a shorter handshake. There's a single Client-Hello and Server-Hello which matches the log above and explains why the SslStream didn't fail when the connection was closed. (I haven't tried parsing out the raw TLS to confirm this yet)
https://cabulous.medium.com/tls-1-2-andtls-1-3-handshake-walkthrough-4cfd0a798164#3252
Next questions:
Thanks for contacting us.
We're moving this issue to the Next sprint planning
milestone for future evaluation / consideration. We would like to keep this around to collect more feedback, which can help us with prioritizing this work. We will re-evaluate this issue, during our next planning meeting(s).
If we later determine, that the issue has no community involvement, or it's very rare and low-impact issue, we will close it - so that the team can focus on more important and high impact issues.
To learn more about what to expect next and how this issue will be handled you can read more about our triage process here.
Hasn't failed in 30 days.
Hasn't failed in 30 days.
Do you happen to know if there is a product change that resulted in the tests passing?
Nope, but a lot has changed in two years, both in kestrel and in underlying layers.
Nope, but a lot has changed in two years, both in kestrel and in underlying layers.
I see. I believe we've recently adopted a pattern of unquarantining things after a test-fixed PR has gone out but considering the age of this one it might be fine to unquarantine and see if they come back.
Failing Test(s)
Microsoft.AspNetCore.Server.Kestrel.InMemory.FunctionalTests.LoggingConnectionMiddlewareTests.LoggingConnectionMiddlewareCanBeAddedBeforeAndAfterHttps
Error Message
Stacktrace
Logs
Build