Feature limit caching to prescribed number of bytes per file (#6009)

* feature: we can check if a fileId is already in the cache

We using this to protect cache from adding the same needle to
the cache over and over.

* fuse mount: Do not start dowloader if needle is already in the cache

* added maxFilePartSizeInCache property to ChunkCache

If file very large only first maxFilePartSizeInCache bytes
are going to be put to the cache (subject to the needle size
constrains).

* feature: for large files put in cache no more than prescribed number of bytes

Before this patch only the first needle of a large file was intended for
caching. This patch uses maximum prescribed amount of bytes to be put in
cache. This allows to bypass default 2MB maximum for a file part stored
in the cache.

* added dummy mock methods to satisfy interfaces of ChunkCache
This commit is contained in:
Eugeniy E. Mikhailov
2024-09-12 00:09:20 -04:00
committed by GitHub
parent 151f2ff7a9
commit dab0bb8097
4 changed files with 61 additions and 2 deletions

View File

@@ -199,7 +199,7 @@ func (c *ChunkReadAt) readChunkSliceAt(buffer []byte, chunkView *ChunkView, next
return fetchChunkRange(buffer, c.readerCache.lookupFileIdFn, chunkView.FileId, chunkView.CipherKey, chunkView.IsGzipped, int64(offset))
}
n, err = c.readerCache.ReadChunkAt(buffer, chunkView.FileId, chunkView.CipherKey, chunkView.IsGzipped, int64(offset), int(chunkView.ChunkSize), chunkView.ViewOffset == 0)
n, err = c.readerCache.ReadChunkAt(buffer, chunkView.FileId, chunkView.CipherKey, chunkView.IsGzipped, int64(offset), int(chunkView.ChunkSize), (uint64(chunkView.ViewOffset)+chunkView.ChunkSize) <= c.readerCache.chunkCache.GetMaxFilePartSizeInCache())
if c.lastChunkFid != chunkView.FileId {
if chunkView.OffsetInChunk == 0 { // start of a new chunk
if c.lastChunkFid != "" {