Use special sentinel for default & handle set 0 differently for direc…

… and non-direct IO

db/db_impl/db_impl_open.cc

@@ -143,11 +143,12 @@ DBOptions SanitizeOptions(const std::string& dbname, const DBOptions& src,
     result.wal_dir = result.wal_dir.substr(0, result.wal_dir.size() - 1);
   }
 
-  if (result.compaction_readahead_size == 0) {
-    if (result.use_direct_reads) {
+  if (result.compaction_readahead_size == Options().compaction_readahead_size ||
+      (result.use_direct_reads && result.compaction_readahead_size == 0)) {
+    if (result.use_direct_reads && result.compaction_readahead_size == 0) {
       TEST_SYNC_POINT_CALLBACK("SanitizeOptions:direct_io", nullptr);
     }
-    result.compaction_readahead_size = 1024 * 1024 * 2;
+    result.compaction_readahead_size = 2 * 1024 * 1024;
   }
 
   // Force flush on DB open if 2PC is enabled, since with 2PC we have no

db/db_options_test.cc

@@ -1034,30 +1034,40 @@ TEST_F(DBOptionsTest, SetFIFOCompactionOptions) {
 }
 
 TEST_F(DBOptionsTest, CompactionReadaheadSizeChange) {
-  SpecialEnv env(env_);
-  Options options;
-  options.env = &env;
-
-  options.compaction_readahead_size = 0;
-  options.level0_file_num_compaction_trigger = 2;
   const std::string kValue(1024, 'v');
-  Reopen(options);
-
-  ASSERT_EQ(1024 * 1024 * 2,
-            dbfull()->GetDBOptions().compaction_readahead_size);
-  ASSERT_OK(dbfull()->SetDBOptions({{"compaction_readahead_size", "256"}}));
-  ASSERT_EQ(256, dbfull()->GetDBOptions().compaction_readahead_size);
-  for (int i = 0; i < 1024; i++) {
-    ASSERT_OK(Put(Key(i), kValue));
-  }
-  ASSERT_OK(Flush());
-  for (int i = 0; i < 1024 * 2; i++) {
-    ASSERT_OK(Put(Key(i), kValue));
+  for (bool direct_io : {true, false}) {
+    SpecialEnv env(env_);
+    Options options;
+    options.env = &env;
+    options.level0_file_num_compaction_trigger = 2;
+    options.use_direct_reads = direct_io;
+    Reopen(options);
+    // Verify the default value of `Option::compaction_readahead_size`
+    ASSERT_EQ(2 * 1024 * 1024,
+              dbfull()->GetDBOptions().compaction_readahead_size);
+    Close();
+    options.compaction_readahead_size = 0;
+    Reopen(options);
+    // Verify the effect of setting `Option::compaction_readahead_size` to be
+    // zero under direct IO and non-Direct IO
+    ASSERT_EQ(direct_io ? 2 * 1024 * 1024 : 0,
+              dbfull()->GetDBOptions().compaction_readahead_size);
+    // Verify `Option::compaction_readahead_size` is set dynamically to positive
+    // value correctly
+    ASSERT_OK(dbfull()->SetDBOptions({{"compaction_readahead_size", "256"}}));
+    ASSERT_EQ(256, dbfull()->GetDBOptions().compaction_readahead_size);
+    for (int i = 0; i < 1024; i++) {
+      ASSERT_OK(Put(Key(i), kValue));
+    }
+    ASSERT_OK(Flush());
+    for (int i = 0; i < 1024 * 2; i++) {
+      ASSERT_OK(Put(Key(i), kValue));
+    }
+    ASSERT_OK(Flush());
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    ASSERT_EQ(256, env_->compaction_readahead_size_);
+    Close();
   }
-  ASSERT_OK(Flush());
-  ASSERT_OK(dbfull()->TEST_WaitForCompact());
-  ASSERT_EQ(256, env_->compaction_readahead_size_);
-  Close();
 }
 
 TEST_F(DBOptionsTest, FIFOTtlBackwardCompatible) {

include/rocksdb/options.h

@@ -952,16 +952,19 @@ struct DBOptions {
   AccessHint access_hint_on_compaction_start = NORMAL;
 
   // The size RocksDB uses to perform readahead during compaction read.
-  // If set zero, RocksDB will sanitize it to be 2MB during db open.
   // If you're
   // running RocksDB on spinning disks, you should set this to at least 2MB.
   // That way RocksDB's compaction is doing sequential instead of random reads.
   //
   //
-  // Default: 0
+  // Default: 2MB
+  //
+  // For direct IO, if set 0, RocksDB will sanitize the value to the default.
+  // For non-direct IO, if set 0, RocksDB will disable compaction readahead
+  // instead.
   //
   // Dynamically changeable through SetDBOptions() API.
-  size_t compaction_readahead_size = 0;
+  size_t compaction_readahead_size = 0xfffffffffffffffe;
 
   // This is a maximum buffer size that is used by WinMmapReadableFile in
   // unbuffered disk I/O mode. We need to maintain an aligned buffer for

unreleased_history/behavior_changes/compaction_readahead_size_option_change.md

@@ -0,0 +1 @@
+`Options::compaction_readahead_size` 's default value is changed from 0 to 2MB. If it's explictly set 0, for direct IO, RocksDB will sanitize the value to the default, while for non-direct IO, RocksDB will disable compaction readahead instead.