Move min_bytes_auto_zstd_compression to compression.go

enterprise/server/backends/migration_cache/migration_cache.go

@@ -1025,14 +1025,13 @@ func (mc *MigrationCache) copy(c *copyData) {
 		log.CtxDebugf(ctx, "Migration already copied on dequeue, returning early: digest %v", c.d)
 		return
 	}
-	if c.d.GetDigest().GetSizeBytes() > 100 && c.conf.src.SupportsCompressor(repb.Compressor_ZSTD) && c.conf.dest.SupportsCompressor(repb.Compressor_ZSTD) {
+	if c.d.GetDigest().GetSizeBytes() >= *compression.MinBytesAutoZstdCompression && c.conf.src.SupportsCompressor(repb.Compressor_ZSTD) && c.conf.dest.SupportsCompressor(repb.Compressor_ZSTD) {
 		// Use compression if both caches support it. This will usually mean
 		// that at src cache doesn't need to decompress and the dest cache
 		// doesn't need to compress, which saves CPU and speeds up the copy.
 		// Ideally, we would only do this when the destination cache would
-		// automatically compress, but since there's no way to check that, rely
-		// on the fact that `cache.pebble.min_bytes_auto_zstd_compression` is
-		// 100.
+		// automatically compress, but since there's no way to check that, check
+		// the same flag that the destination cache would use.
 		c.d = c.d.CloneVT()
 		c.d.Compressor = repb.Compressor_ZSTD
 	}

enterprise/server/backends/pebble_cache/pebble_cache.go

@@ -95,9 +95,6 @@ var (
 	activeKeyVersion  = flag.Int64("cache.pebble.active_key_version", int64(filestore.UnspecifiedKeyVersion), "The key version new data will be written with. If negative, will write to the highest existing version in the database, or the highest known version if a new database is created.")
 	migrationQPSLimit = flag.Int("cache.pebble.migration_qps_limit", 50, "QPS limit for data version migration")
 
-	// Compression related flags
-	minBytesAutoZstdCompression = flag.Int64("cache.pebble.min_bytes_auto_zstd_compression", 100, "Blobs larger than this will be zstd compressed before written to disk.")
-
 	// Chunking related flags
 	averageChunkSizeBytes = flag.Int("cache.pebble.average_chunk_size_bytes", 0, "Average size of chunks that's stored in the cache. Disabled if 0.")
 
@@ -342,7 +339,7 @@ func Register(env *real_environment.RealEnv) error {
 		BlockCacheSizeBytes:         *blockCacheSizeBytesFlag,
 		MaxSizeBytes:                cache_config.MaxSizeBytes(),
 		MaxInlineFileSizeBytes:      *maxInlineFileSizeBytesFlag,
-		MinBytesAutoZstdCompression: *minBytesAutoZstdCompression,
+		MinBytesAutoZstdCompression: *compression.MinBytesAutoZstdCompression,
 		AtimeUpdateThreshold:        atimeUpdateThresholdFlag,
 		AtimeBufferSize:             atimeBufferSizeFlag,
 		NumAtimeUpdateWorkers:       numAtimeUpdateWorkers,

server/remote_cache/byte_stream_server/byte_stream_server.go

@@ -285,7 +285,7 @@ func (s *ByteStreamServer) beginWrite(ctx context.Context, req *bspb.WriteReques
 		casRN.SetCompressor(r.GetCompressor())
 	}
 	compressData := false
-	if casRN.GetCompressor() == repb.Compressor_IDENTITY && s.cache.SupportsCompressor(repb.Compressor_ZSTD) && r.GetDigest().GetSizeBytes() >= 100 {
+	if casRN.GetCompressor() == repb.Compressor_IDENTITY && s.cache.SupportsCompressor(repb.Compressor_ZSTD) && r.GetDigest().GetSizeBytes() >= *compression.MinBytesAutoZstdCompression {
 		casRN.SetCompressor(repb.Compressor_ZSTD)
 		compressData = true
 	}

server/util/compression/compression.go

@@ -2,6 +2,7 @@ package compression
 
 import (
 	"errors"
+	"flag"
 	"io"
 	"runtime"
 	"sync"
@@ -31,6 +32,10 @@ var (
 	zstdDecompressedBytesMetric = metrics.BytesDecompressed.With(prometheus.Labels{metrics.CompressionType: "zstd"})
 )
 
+var (
+	MinBytesAutoZstdCompression = flag.Int64("compression.min_bytes_auto_zstd_compression", 100, "Blobs larger than this will be zstd compressed before written to disk.")
+)
+
 func mustGetZstdEncoder() *zstd.Encoder {
 	enc, err := zstd.NewWriter(nil)
 	if err != nil {