tvl-depot/tvix/nar-bridge/pkg/importer/importer.go
Connor Brewster 85421b7f97 feat(tvix/store/nar-info): Support async blob upload
Right now all blob uploads are performed synchronously, this means if a
NAR contains many small files, the import time is dominated by round
trip time to the blob service. For small files, we can buffer them in
memory and upload them asynchronously to the blob service. Before
returning we make sure to join all the uploads to make sure they
complete successfully before responding OK.

This reduces time to import a bash-interactive closure on my machine
from 1m19s to 7s.

Change-Id: Ica3695c159e6c8ad8769281ac20d037e3143e856
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10679
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: Connor Brewster <cbrewster@hey.com>
2024-01-23 17:01:16 +00:00

303 lines
9.6 KiB
Go

package importer
import (
"bytes"
"context"
"crypto/sha256"
"errors"
"fmt"
"io"
"path"
"strings"
castorev1pb "code.tvl.fyi/tvix/castore-go"
"github.com/nix-community/go-nix/pkg/nar"
"golang.org/x/sync/errgroup"
"lukechampine.com/blake3"
)
const (
// asyncUploadThreshold controls when a file is buffered into memory and uploaded
// asynchronously. Files must be smaller than the threshold to be uploaded asynchronously.
asyncUploadThreshold = 1024 * 1024 // 1 MiB
// maxAsyncUploadBufferBytes is the maximum number of async blob uploads allowed to be
// running concurrently at any given time for a simple import operation.
maxConcurrentAsyncUploads = 128
)
// An item on the directories stack
type stackItem struct {
path string
directory *castorev1pb.Directory
}
// Import reads a NAR from a reader, and returns a the root node,
// NAR size and NAR sha256 digest.
func Import(
// a context, to support cancellation
ctx context.Context,
// The reader the data is read from
r io.Reader,
// callback function called with each regular file content
blobCb func(fileReader io.Reader) ([]byte, error),
// callback function called with each finalized directory node
directoryCb func(directory *castorev1pb.Directory) ([]byte, error),
) (*castorev1pb.Node, uint64, []byte, error) {
// We need to wrap the underlying reader a bit.
// - we want to keep track of the number of bytes read in total
// - we calculate the sha256 digest over all data read
// Express these two things in a MultiWriter, and give the NAR reader a
// TeeReader that writes to it.
narCountW := &CountingWriter{}
sha256W := sha256.New()
multiW := io.MultiWriter(narCountW, sha256W)
narReader, err := nar.NewReader(io.TeeReader(r, multiW))
if err != nil {
return nil, 0, nil, fmt.Errorf("failed to instantiate nar reader: %w", err)
}
defer narReader.Close()
// If we store a symlink or regular file at the root, these are not nil.
// If they are nil, we instead have a stackDirectory.
var rootSymlink *castorev1pb.SymlinkNode
var rootFile *castorev1pb.FileNode
var stackDirectory *castorev1pb.Directory
// Keep track of all asynch blob uploads so we can make sure they all succeed
// before returning.
var asyncBlobWg errgroup.Group
asyncBlobWg.SetLimit(maxConcurrentAsyncUploads)
var stack = []stackItem{}
// popFromStack is used when we transition to a different directory or
// drain the stack when we reach the end of the NAR.
// It adds the popped element to the element underneath if any,
// and passes it to the directoryCb callback.
// This function may only be called if the stack is not already empty.
popFromStack := func() error {
// Keep the top item, and "resize" the stack slice.
// This will only make the last element unaccessible, but chances are high
// we're re-using that space anyways.
toPop := stack[len(stack)-1]
stack = stack[:len(stack)-1]
// call the directoryCb
directoryDigest, err := directoryCb(toPop.directory)
if err != nil {
return fmt.Errorf("failed calling directoryCb: %w", err)
}
// if there's still a parent left on the stack, refer to it from there.
if len(stack) > 0 {
topOfStack := stack[len(stack)-1].directory
topOfStack.Directories = append(topOfStack.Directories, &castorev1pb.DirectoryNode{
Name: []byte(path.Base(toPop.path)),
Digest: directoryDigest,
Size: toPop.directory.Size(),
})
}
// Keep track that we have encounter at least one directory
stackDirectory = toPop.directory
return nil
}
getBasename := func(p string) string {
// extract the basename. In case of "/", replace with empty string.
basename := path.Base(p)
if basename == "/" {
basename = ""
}
return basename
}
for {
select {
case <-ctx.Done():
return nil, 0, nil, ctx.Err()
default:
// call narReader.Next() to get the next element
hdr, err := narReader.Next()
// If this returns an error, it's either EOF (when we're done reading from the NAR),
// or another error.
if err != nil {
// if this returns no EOF, bail out
if !errors.Is(err, io.EOF) {
return nil, 0, nil, fmt.Errorf("failed getting next nar element: %w", err)
}
// The NAR has been read all the way to the end…
// Make sure we close the nar reader, which might read some final trailers.
if err := narReader.Close(); err != nil {
return nil, 0, nil, fmt.Errorf("unable to close nar reader: %w", err)
}
// Check the stack. While it's not empty, we need to pop things off the stack.
for len(stack) > 0 {
err := popFromStack()
if err != nil {
return nil, 0, nil, fmt.Errorf("unable to pop from stack: %w", err)
}
}
// Wait for any pending blob uploads.
err := asyncBlobWg.Wait()
if err != nil {
return nil, 0, nil, fmt.Errorf("async blob upload: %w", err)
}
// Stack is empty.
// Now either root{File,Symlink,Directory} is not nil,
// and we can return the root node.
narSize := narCountW.BytesWritten()
narSha256 := sha256W.Sum(nil)
if rootFile != nil {
return &castorev1pb.Node{
Node: &castorev1pb.Node_File{
File: rootFile,
},
}, narSize, narSha256, nil
} else if rootSymlink != nil {
return &castorev1pb.Node{
Node: &castorev1pb.Node_Symlink{
Symlink: rootSymlink,
},
}, narSize, narSha256, nil
} else if stackDirectory != nil {
// calculate directory digest (i.e. after we received all its contents)
dgst, err := stackDirectory.Digest()
if err != nil {
return nil, 0, nil, fmt.Errorf("unable to calculate root directory digest: %w", err)
}
return &castorev1pb.Node{
Node: &castorev1pb.Node_Directory{
Directory: &castorev1pb.DirectoryNode{
Name: []byte{},
Digest: dgst,
Size: stackDirectory.Size(),
},
},
}, narSize, narSha256, nil
} else {
return nil, 0, nil, fmt.Errorf("no root set")
}
}
// Check for valid path transitions, pop from stack if needed
// The nar reader already gives us some guarantees about ordering and illegal transitions,
// So we really only need to check if the top-of-stack path is a prefix of the path,
// and if it's not, pop from the stack. We do this repeatedly until the top of the stack is
// the subdirectory the new entry is in, or we hit the root directory.
// We don't need to worry about the root node case, because we can only finish the root "/"
// If we're at the end of the NAR reader (covered by the EOF check)
for len(stack) > 1 && !strings.HasPrefix(hdr.Path, stack[len(stack)-1].path+"/") {
err := popFromStack()
if err != nil {
return nil, 0, nil, fmt.Errorf("unable to pop from stack: %w", err)
}
}
if hdr.Type == nar.TypeSymlink {
symlinkNode := &castorev1pb.SymlinkNode{
Name: []byte(getBasename(hdr.Path)),
Target: []byte(hdr.LinkTarget),
}
if len(stack) > 0 {
topOfStack := stack[len(stack)-1].directory
topOfStack.Symlinks = append(topOfStack.Symlinks, symlinkNode)
} else {
rootSymlink = symlinkNode
}
}
if hdr.Type == nar.TypeRegular {
uploadBlob := func(r io.Reader) ([]byte, error) {
// wrap reader with a reader counting the number of bytes read
blobCountW := &CountingWriter{}
blobReader := io.TeeReader(r, blobCountW)
blobDigest, err := blobCb(blobReader)
if err != nil {
return nil, fmt.Errorf("failure from blobCb: %w", err)
}
// ensure blobCb did read all the way to the end.
// If it didn't, the blobCb function is wrong and we should bail out.
if blobCountW.BytesWritten() != uint64(hdr.Size) {
return nil, fmt.Errorf("blobCb did not read all: %d/%d bytes", blobCountW.BytesWritten(), hdr.Size)
}
return blobDigest, nil
}
h := blake3.New(32, nil)
blobReader := io.TeeReader(narReader, io.MultiWriter(h))
var blobDigest []byte
// If this file is small enough, read it off the wire immediately and
// upload to the blob service asynchronously. This helps reduce the
// RTT on blob uploads for NARs with many small files.
doAsync := hdr.Size < asyncUploadThreshold
if doAsync {
blobContents, err := io.ReadAll(blobReader)
if err != nil {
return nil, 0, nil, fmt.Errorf("read blob: %w", err)
}
blobDigest = h.Sum(nil)
asyncBlobWg.Go(func() error {
blobDigestFromCb, err := uploadBlob(bytes.NewReader(blobContents))
if err != nil {
return err
}
if !bytes.Equal(blobDigest, blobDigestFromCb) {
return fmt.Errorf("unexpected digest (got %x, expected %x)", blobDigestFromCb, blobDigest)
}
return nil
})
} else {
blobDigestFromCb, err := uploadBlob(blobReader)
if err != nil {
return nil, 0, nil, fmt.Errorf("upload blob: %w", err)
}
blobDigest = h.Sum(nil)
if !bytes.Equal(blobDigest, blobDigestFromCb) {
return nil, 0, nil, fmt.Errorf("unexpected digest (got %x, expected %x)", blobDigestFromCb, blobDigest)
}
}
fileNode := &castorev1pb.FileNode{
Name: []byte(getBasename(hdr.Path)),
Digest: blobDigest,
Size: uint64(hdr.Size),
Executable: hdr.Executable,
}
if len(stack) > 0 {
topOfStack := stack[len(stack)-1].directory
topOfStack.Files = append(topOfStack.Files, fileNode)
} else {
rootFile = fileNode
}
}
if hdr.Type == nar.TypeDirectory {
directory := &castorev1pb.Directory{
Directories: []*castorev1pb.DirectoryNode{},
Files: []*castorev1pb.FileNode{},
Symlinks: []*castorev1pb.SymlinkNode{},
}
stack = append(stack, stackItem{
directory: directory,
path: hdr.Path,
})
}
}
}
}