sneak/vaultik
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Read chunks from cached blobs via ReadAt instead of full-blob Get

Browse Source 
Restore's per-chunk loop called blobCache.Get(blobHash) and sliced the
returned []byte to extract the chunk it actually wanted. Get reads the
entire blob from disk into memory — so for a 10 GB blob, every chunk
extraction was a 10 GB ReadFile to get back a few KB. With ~40k files
needing ~600ms per cache hit, that alone was burning ~6 hours of
wall-clock on a real restore.

Hot loop now:
  - If the blob isn't cached: download (full plaintext into memory),
    Put to disk cache, satisfy this chunk from the in-memory buffer.
  - If it's cached: blobCache.ReadAt(hash, offset, length) — reads
    only the chunk's bytes from the on-disk blob file.

ReadAt was already implemented on blobDiskCache; restore just wasn't
using it.

Debug timings from the user's photo-catalog restore showed
ms_cache_gets dominating every cache-hit file at 500-1000ms. With
ReadAt those should drop to sub-millisecond and the visible throughput
should be bound by single-stream blob download + decrypt/decompress
rather than disk-read amplification.
This commit is contained in:
Jeffrey Paul
2026-06-17 07:58:01 +02:00
parent b55d5763ad
commit 8036d93914
1 changed files with 24 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
internal/vaultik/restore.go
View File

@@ -569,14 +569,22 @@ func (v *Vaultik) restoreRegularFile(
return fmt.Errorf("getting blob %s: %w", blobChunk.BlobID, err)
}
// Download and decrypt blob if not cached
// If the blob isn't on disk yet, download it (one full readout
// of the plaintext into memory), write it to the cache, and
// satisfy THIS chunk from the in-memory copy. Subsequent files
// needing chunks from the same blob hit the ReadAt branch and
// read only the bytes they actually want — never the whole
// blob — which is the difference between sub-millisecond chunk
// extraction and ~1 s per cache hit on a 10 GB blob.
blobHashStr := blob.Hash.String()
var chunkData []byte
t0 = time.Now()
blobData, ok := blobCache.Get(blobHashStr)
blobInCache := blobCache.Has(blobHashStr)
cacheGetDur += time.Since(t0)
if !ok {
if !blobInCache {
t0 = time.Now()
blobData, err = v.downloadBlob(ctx, blobHashStr, blob.CompressedSize, identity)
blobData, err := v.downloadBlob(ctx, blobHashStr, blob.CompressedSize, identity)
downloadDur += time.Since(t0)
if err != nil {
return fmt.Errorf("downloading blob %s: %w", blobHashStr[:16], err)
@@ -589,16 +597,21 @@ func (v *Vaultik) restoreRegularFile(
downloadCount++
result.BlobsDownloaded++
result.BytesDownloaded += blob.CompressedSize
} else {
cacheHitCount++
}
// Extract chunk from blob
if blobChunk.Offset+blobChunk.Length > int64(len(blobData)) {
return fmt.Errorf("chunk %s extends beyond blob data (offset=%d, length=%d, blob_size=%d)",
fc.ChunkHash[:16], blobChunk.Offset, blobChunk.Length, len(blobData))
}
chunkData := blobData[blobChunk.Offset : blobChunk.Offset+blobChunk.Length]
chunkData = blobData[blobChunk.Offset : blobChunk.Offset+blobChunk.Length]
} else {
cacheHitCount++
t0 = time.Now()
chunkData, err = blobCache.ReadAt(blobHashStr, blobChunk.Offset, blobChunk.Length)
cacheGetDur += time.Since(t0)
if err != nil {
return fmt.Errorf("reading chunk %s from cached blob %s: %w", fc.ChunkHash[:16], blobHashStr[:16], err)
}
}
// Write chunk to output file
t0 = time.Now()