Complete MergeHub Sink gracefully on NormalShutdownException

[anpin]

Fixes #6931

Changes

Pattern match exception to avoid unnecessary errors logged upon Akka.IO streams failure

Checklist

For significant changes, please ensure that the following have been completed (delete if not relevant):

• This change follows the Akka.NET API Compatibility Guidelines.
• This change follows the Akka.NET Wire Compatibility Guidelines.
• I have reviewed my own pull request.
• Design discussion issue IO-TCP logs unnecessary error message on flow termination #6931

Latest dev Benchmarks

Include data from the relevant benchmark prior to this change here.

This PR's Benchmarks

Include data from after this change here.

[anpin]

Draft targets 1.5.30 tag since that is the version in my project. I can rebase to dev once we agree that this should be merged.

[Aaronontheweb]

@anpin changes look good but this spec is failing (please check the build log):

System.AggregateException : One or more errors occurred. (Timeout (00:00:03) while waiting for messages. Only received 0/1 messages that matched filter [Error when Message contains "Upstream producer failed with exception"])
---- Timeout (00:00:03) while waiting for messages. Only received 0/1 messages that matched filter [Error when Message contains "Upstream producer failed with exception"]

Is that to be expected?

[Aaronontheweb]

LGTM but please see my comment RE: test failrues

[Aaronontheweb]

The other two tests that failed:

1. LogFormatSpec should be addressed when we rebase on dev, which I have just done
2. The TCP chain of echoes spec is a frequently racy unit test and is not impacted by this change, I think

[anpin]

Is that to be expected?

I don't know, seems to work fine on my end in Rider. Maybe FAKE on Windows somehow leads to a different result?

[Aaronontheweb]

Give it a "run until failure" and let it cook for a little bit (like 10-15 minutes while you're doing something else) - I had to do that myself today for a couple of specs I was working on.

[Aaronontheweb]

But otherwise you can mark this PR as "ready for review"

[anpin]

But otherwise you can mark this PR as "ready for review"

I''ll give it a second chance to fail tomorrow using your suggestion and will cleanup those comments in tests. Thank you @Aaronontheweb for looking into this!

[Aaronontheweb]

@anpin thank you for your hard work and research on this issue!

[He-Pin]

Nice find, will port this fix to Akka Scala too?

[anpin]

@He-Pin not a scala user myself, but you are welcome to do so if such an issue exists there

[He-Pin]

I will try to port it to pekko, after this been merged, thanks. @anpin

[anpin]

LogFormatSpec should be addressed when we rebase on dev

@Aaronontheweb could you clarify what you meant by this?

I couldn't make those test fail in Rider on Linux. However I spotted a lot of error logs on linux in CI (for all PR's not this one in particular) and that the build script depends on outdated FAKE tooling, so my results are potentially not reproducible due to difference in environment used.

[anpin]

I poked a few of these failing tests locally

[Aaronontheweb]

LogFormatSpec should be addressed when we rebase on dev

@Aaronontheweb could you clarify what you meant by this?

Mistake on my part - I thought there was an issue with the log formatting being broken in the version of Akka.NET (v1.5.30) you'd based your branch on. There might be a race condition in that test (very common and annoying in a test suite as large as ours.)

[Aaronontheweb]

The FAKE build errors is mostly output from https://github.com/petabridge/Incrementalist - which can run into some solution analysis hiccups that aren't significant

[Aaronontheweb]

LGTM