10477: disable sha-256 calculation by the S3 driver; we don't need it

       and it slows uploads down because it's CPU bound.

Arvados-DCO-1.1-Signed-off-by: Ward Vandewege <ward@curii.com>

diff --git a/doc/install/configure-s3-object-storage.html.textile.liquid b/doc/install/configure-s3-object-storage.html.textile.liquid
index cfd436dd6f4b660f383851ae1249ebb50d2886a1..40cbbb533d91989016923cffe3a5242e0fe34815 100644 (file)
--- a/doc/install/configure-s3-object-storage.html.textile.liquid
+++ b/doc/install/configure-s3-object-storage.html.textile.liquid
@@ -107,11 +107,11 @@ The @aws-sdk-go-v2@ driver has faster _single thread_ read and write performance
 table(table table-bordered table-condensed).
 ||_. goamz |_. aws-sdk-go-v2 |_. command line|
 |single thread read performance (average)|32.53 MiB/s|79.48 MiB/s|keep-exercise -repeat 4 -run-time 60s -vary-request -use-index -rthreads 1 -wthreads 1|
-|single thread write performance (average)|39.75 MiB/s|49.58 MiB/s|keep-exercise -repeat 4 -run-time 60s -vary-request -use-index -rthreads 1 -wthreads 1|
+|single thread write performance (average)|39.75 MiB/s|41.05 MiB/s|keep-exercise -repeat 4 -run-time 60s -vary-request -use-index -rthreads 1 -wthreads 1|
 
 Because both S3 and Keep are optimized for _aggregate_ througput, the single thread performance is not as important as it may seem at first glance. When using 20 concurrent read or write threads, the numbers from both drivers are more closely aligned:
 
 table(table table-bordered table-condensed).
 ||_. goamz |_. aws-sdk-go-v2 |_. command line|
 |20 thread read performance (average)|585.60 MiB/s|898.93 MiB/s|keep-exercise -repeat 4 -run-time 60s -vary-request -use-index -rthreads 20 -wthreads 0|
-|20 thread write performance (average)|610.40 MiB/s|542.40 MiB/s|keep-exercise -repeat 4 -run-time 60s -vary-request -use-index -rthreads 0 -wthreads 20|
+|20 thread write performance (average)|610.40 MiB/s|688.25 MiB/s|keep-exercise -repeat 4 -run-time 60s -vary-request -use-index -rthreads 0 -wthreads 20|

diff --git a/services/keepstore/s3aws_volume.go b/services/keepstore/s3aws_volume.go
index d4fe9fd156032d95868319c23c110da5ff0b2831..9bbb3c5b1c43f103e3cf2eb6cad997703ccde75d 100644 (file)
--- a/services/keepstore/s3aws_volume.go
+++ b/services/keepstore/s3aws_volume.go
@@ -534,7 +534,9 @@ func (b *s3AWSbucket) PutReader(path string, r io.Reader, length int64, contType
                Bucket: aws.String(b.bucket),
                Key:    aws.String(path),
                Body:   r,
-       })
+       }, s3manager.WithUploaderRequestOptions(func(r *aws.Request) {
+               r.HTTPRequest.Header.Set("X-Amz-Content-Sha256", "UNSIGNED-PAYLOAD")
+       }))
 
        b.stats.TickOps("put")
        b.stats.Tick(&b.stats.Ops, &b.stats.PutOps)
@@ -570,10 +572,6 @@ func (v *S3AWSVolume) WriteBlock(ctx context.Context, loc string, rdr io.Reader)
        // See if this is the empty block
        if contentMD5 != "d41d8cd98f00b204e9800998ecf8427e" {
                uploadInput.ContentMD5 = &contentMD5
-               // Unlike the goamz S3 driver, we don't need to precompute ContentSHA256:
-               // the aws-sdk-go v2 SDK uses a ReadSeeker to avoid having to copy the
-               // block, so there is no extra memory use to be concerned about. See
-               // makeSha256Reader in aws/signer/v4/v4.go.
        }
 
        // Some experimentation indicated that using concurrency 5 yields the best
@@ -585,7 +583,15 @@ func (v *S3AWSVolume) WriteBlock(ctx context.Context, loc string, rdr io.Reader)
                u.Concurrency = 5
        })
 
-       _, err = uploader.UploadWithContext(ctx, &uploadInput, s3manager.WithUploaderRequestOptions())
+       // Unlike the goamz S3 driver, we don't need to precompute ContentSHA256:
+       // the aws-sdk-go v2 SDK uses a ReadSeeker to avoid having to copy the
+       // block, so there is no extra memory use to be concerned about. See
+       // makeSha256Reader in aws/signer/v4/v4.go. In fact, we explicitly disable
+       // calculating the Sha-256 because we don't need it; we already use md5sum
+       // hashes that match the name of the block.
+       _, err = uploader.UploadWithContext(ctx, &uploadInput, s3manager.WithUploaderRequestOptions(func(r *aws.Request) {
+               r.HTTPRequest.Header.Set("X-Amz-Content-Sha256", "UNSIGNED-PAYLOAD")
+       }))
 
        v.bucket.stats.TickOps("put")
        v.bucket.stats.Tick(&v.bucket.stats.Ops, &v.bucket.stats.PutOps)
@@ -599,7 +605,9 @@ func (v *S3AWSVolume) WriteBlock(ctx context.Context, loc string, rdr io.Reader)
                Bucket: aws.String(v.bucket.bucket),
                Key:    aws.String("recent/" + loc),
                Body:   empty,
-       })
+       }, s3manager.WithUploaderRequestOptions(func(r *aws.Request) {
+               r.HTTPRequest.Header.Set("X-Amz-Content-Sha256", "UNSIGNED-PAYLOAD")
+       }))
        v.bucket.stats.TickOps("put")
        v.bucket.stats.Tick(&v.bucket.stats.Ops, &v.bucket.stats.PutOps)
        v.bucket.stats.TickErr(err)