AwsSdkPipelineWrapper: Asynchronous task blocked on a synchronous call

[gjunge]

Description
After upgrading to 10.31.0 from 10.26.0 we have seen on a number of occasions that the CPU spiked to almost 100% and stayed there. After windbg-ing we have seen that we have a race condition in the new AwsSdkPipelineWrapper when inserting a new value into the HashSet<string> _unsupportedRequestTypes.
In this one instance we had around 8 threads locked on this call.

Steps to Reproduce
This happened when we have a [Trace] attribute on a asynchronous method that lists objects from S3.

        [Trace]
        public async Task<IEnumerable<S3Object>> GetFilesInternalAsync(string path)
        {
             var results = new List<S3Object>();
            ListObjectsV2Response s3Response = null;
            do
            {
                s3Response = await _s3client.ListObjectsV2Async(new ListObjectsV2Request() {
                    Prefix = path,
                    BucketName = this.Bucket,
                    ContinuationToken = s3Response?.NextContinuationToken
                }).ConfigureAwait(false);

                results.AddRange(s3Response.S3Objects.Where(f => f.Key != path));

            } while (s3Response.IsTruncated);

            return results;
        }

Your Environment
.NET 4.8.2 , Agent 10.31.0

[workato-integration]

https://new-relic.atlassian.net/browse/NR-333108

[tippmar-nr]

@gjunge Thanks for reporting this issue and for submitting a PR to fix it.

We are reviewing your PR and will likely accept the change via a PR directly on this repo - we need to be able to run our full set of CI workflows, which unfortunately can't run on a fork.

[gjunge]

Thanks for the quick response!

[wr-cdargis]

Just came on here to report this. So glad it's already addressed.

When will this fix be released? @gjunge @tippmar-nr ?

[nrcventura]

It was included in the 10.33.0 release.