LevelDB now attempts to reuse the preceding MANIFEST and log file when re-opened.

(Based on a suggestion by cmumford.)

"open" benchmark on my workstation speeds up significantly since we
can now avoid three fdatasync calls and a compaction per open:

  Before: ~80000 microseconds
  After:    ~130 microseconds

Details:

(1) Added Options::reuse_logs (currently defaults to false) to control
new behavior.  The intention is to change the default to true after some
baking.

(2) Added Env::NewAppendableFile() whose default implementation returns
a not-supported error.

(3) VersionSet::Recovery attempts to reuse the MANIFEST from which
it is recovering.

(4) DBImpl recovery attempts to reuse the last log file and memtable.

(5) db_test.cc now tests a new configuration that sets reuse_logs to true.

(6) fault_injection_test also tests a reuse_logs==true config.

(7) Added a new recovery_test.

1 files changed, 20 insertions, 1 deletions

diff --git a/db/db_test.cc b/db/db_test.cc
index 0fed913..74ba8b0 100644
--- a/db/db_test.cc
+++ b/db/db_test.cc
@@ -193,6 +193,7 @@ class DBTest {
   // Sequence of option configurations to try
   enum OptionConfig {
     kDefault,
+    kReuse,
     kFilter,
     kUncompressed,
     kEnd
@@ -237,7 +238,11 @@ class DBTest {
   // Return the current option configuration.
   Options CurrentOptions() {
     Options options;
+    options.reuse_logs = false;
     switch (option_config_) {
+      case kReuse:
+        options.reuse_logs = true;
+        break;
       case kFilter:
         options.filter_policy = filter_policy_;
         break;
@@ -1080,6 +1085,14 @@ TEST(DBTest, ApproximateSizes) {
     // 0 because GetApproximateSizes() does not account for memtable space
     ASSERT_TRUE(Between(Size("", Key(50)), 0, 0));
 
+    if (options.reuse_logs) {
+      // Recovery will reuse memtable, and GetApproximateSizes() does not
+      // account for memtable usage;
+      Reopen(&options);
+      ASSERT_TRUE(Between(Size("", Key(50)), 0, 0));
+      continue;
+    }
+
     // Check sizes across recovery by reopening a few times
     for (int run = 0; run < 3; run++) {
       Reopen(&options);
@@ -1123,6 +1136,11 @@ TEST(DBTest, ApproximateSizes_MixOfSmallAndLarge) {
     ASSERT_OK(Put(Key(6), RandomString(&rnd, 300000)));
     ASSERT_OK(Put(Key(7), RandomString(&rnd, 10000)));
 
+    if (options.reuse_logs) {
+      // Need to force a memtable compaction since recovery does not do so.
+      ASSERT_OK(dbfull()->TEST_CompactMemTable());
+    }
+
     // Check sizes across recovery by reopening a few times
     for (int run = 0; run < 3; run++) {
       Reopen(&options);
@@ -2084,7 +2102,8 @@ void BM_LogAndApply(int iters, int num_base_files) {
   InternalKeyComparator cmp(BytewiseComparator());
   Options options;
   VersionSet vset(dbname, &options, NULL, &cmp);
-  ASSERT_OK(vset.Recover());
+  bool save_manifest;
+  ASSERT_OK(vset.Recover(&save_manifest));
   VersionEdit vbase;
   uint64_t fnum = 1;
   for (int i = 0; i < num_base_files; i++) {