Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

fix(snapshots): choose numCPU if less than max parallel value #3069

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

fix(snapshots): choose numCPU if less than max parallel value #3069

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3960345
Kopia parallelism: choose numCPU if less than max parallel value
Jun 14, 2023
036e042
Merge branch 'master' into snapshots/parallel-cpu-num
MinyiZou Jun 14, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions snapshot/snapshotfs/upload.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -744,6 +744,11 @@ func (u *Uploader) maybeIgnoreCachedEntry(ctx context.Context, ent fs.Entry) fs.

func (u *Uploader) effectiveParallelFileReads(pol *policy.Policy) int {
p := u.ParallelUploads

if p > runtime.NumCPU() {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

given that uploads are not entirely CPU bound there may be benefits to allowing p to be greater than the number of cores to allow max number of cores to do hashing while others are waiting for uploads

p = runtime.NumCPU()
}

if p > 0 {
// command-line override takes precedence.
return p
Expand Down
65 changes: 63 additions & 2 deletions snapshot/snapshotfs/upload_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,7 @@ import (
"os"
"path/filepath"
"reflect"
"runtime"
"runtime/debug"
"sort"
"strings"
Expand Down Expand Up @@ -688,7 +689,7 @@ func TestUploadWithCheckpointing(t *testing.T) {
}
}

func TestParallelUploadUploadsBlobsInParallel(t *testing.T) {
func TestParallelUploadUploadsBlobsInParallelOverNumCPU(t *testing.T) {
ctx := testlogging.Context(t)
th := newUploadTestHarness(ctx, t)

Expand Down Expand Up @@ -718,7 +719,67 @@ func TestParallelUploadUploadsBlobsInParallel(t *testing.T) {

policyTree := policy.BuildTree(nil, policy.DefaultPolicy)

require.Equal(t, 13, u.effectiveParallelFileReads(policyTree.EffectivePolicy()))
n := runtime.NumCPU()

require.Equal(t, n, u.effectiveParallelFileReads(policyTree.EffectivePolicy()))

si := snapshot.SourceInfo{
UserName: "user",
Host: "host",
Path: "path",
}

// add a bunch of very large files which can be hashed in parallel and will trigger parallel
// uploads
th.sourceDir.AddFile("d1/large1", randomBytes(1e7), defaultPermissions)
th.sourceDir.AddFile("d1/large2", randomBytes(2e7), defaultPermissions)
th.sourceDir.AddFile("d1/large3", randomBytes(2e7), defaultPermissions)
th.sourceDir.AddFile("d1/large4", randomBytes(1e7), defaultPermissions)

th.sourceDir.AddFile("d2/large1", randomBytes(1e7), defaultPermissions)
th.sourceDir.AddFile("d2/large2", randomBytes(1e7), defaultPermissions)
th.sourceDir.AddFile("d2/large3", randomBytes(1e7), defaultPermissions)
th.sourceDir.AddFile("d2/large4", randomBytes(1e7), defaultPermissions)

_, err := u.Upload(ctx, th.sourceDir, policyTree, si)
require.NoError(t, err)

require.NoError(t, th.repo.Flush(ctx))

require.Greater(t, maxParallelCalls.Load(), int32(0))
}

func TestParallelUploadUploadsBlobsInParallelWithinNumCPU(t *testing.T) {
ctx := testlogging.Context(t)
th := newUploadTestHarness(ctx, t)

u := NewUploader(th.repo)
u.ParallelUploads = 2

// no faults for first blob write - session marker.
th.faulty.AddFault(blobtesting.MethodPutBlob)

var currentParallelCalls, maxParallelCalls atomic.Int32

// measure concurrency of PutBlob calls
th.faulty.AddFault(blobtesting.MethodPutBlob).Repeat(10).Before(func() {
v := currentParallelCalls.Add(1)
max := maxParallelCalls.Load()
if v > max {
maxParallelCalls.CompareAndSwap(max, v)
}

time.Sleep(100 * time.Millisecond)
currentParallelCalls.Add(-1)
})

// create a channel that will be sent to whenever checkpoint completes.
u.checkpointFinished = make(chan struct{})
u.disableEstimation = true

policyTree := policy.BuildTree(nil, policy.DefaultPolicy)

require.Equal(t, 2, u.effectiveParallelFileReads(policyTree.EffectivePolicy()))

si := snapshot.SourceInfo{
UserName: "user",
Expand Down