Fixed a file-not-found issue when a log file is moved to archive.

Summary:
Fixed a file-not-found issue when a log file is moved to archive
by doing a missing retry.

Test Plan:
make db_test
export ROCKSDB_TEST=TransactionLogIteratorRace
./db_test

Reviewers: sdong, haobo

Reviewed By: sdong

CC: igor, leveldb

Differential Revision: https://reviews.facebook.net/D18669

db/db_impl.cc

@@ -935,8 +935,19 @@ Status DBImpl::GetSortedWalsOfType(const std::string& path,
         continue;
       }
 
+      // Reproduce the race condition where a log file is moved
+      // to archived dir, between these two sync points, used in
+      // (DBTest,TransactionLogIteratorRace)
+      TEST_SYNC_POINT("DBImpl::GetSortedWalsOfType:1");
+      TEST_SYNC_POINT("DBImpl::GetSortedWalsOfType:2");
+
       uint64_t size_bytes;
       s = env_->GetFileSize(LogFileName(path, number), &size_bytes);
+      // re-try in case the alive log file has been moved to archive.
+      if (!s.ok() && log_type == kAliveLogFile &&
+          env_->FileExists(ArchivedLogFileName(path, number))) {
+        s = env_->GetFileSize(ArchivedLogFileName(path, number), &size_bytes);
+      }
       if (!s.ok()) {
         return s;
       }

db/db_test.cc

@@ -5609,48 +5609,56 @@ TEST(DBTest, TransactionLogIterator) {
 
 #ifndef NDEBUG // sync point is not included with DNDEBUG build
 TEST(DBTest, TransactionLogIteratorRace) {
-  // Setup sync point dependency to reproduce the race condition of
+  static const int LOG_ITERATOR_RACE_TEST_COUNT = 2;
-  // a log file moved to archived dir, in the middle of GetSortedWalFiles
+  static const char* sync_points[LOG_ITERATOR_RACE_TEST_COUNT][4] =
-  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+    { { "DBImpl::GetSortedWalFiles:1", "DBImpl::PurgeObsoleteFiles:1",
-    { { "DBImpl::GetSortedWalFiles:1", "DBImpl::PurgeObsoleteFiles:1" },
+        "DBImpl::PurgeObsoleteFiles:2", "DBImpl::GetSortedWalFiles:2" },
-      { "DBImpl::PurgeObsoleteFiles:2", "DBImpl::GetSortedWalFiles:2" },
+      { "DBImpl::GetSortedWalsOfType:1", "DBImpl::PurgeObsoleteFiles:1",
-    });
+        "DBImpl::PurgeObsoleteFiles:2", "DBImpl::GetSortedWalsOfType:2" }};
-
+  for (int test = 0; test < LOG_ITERATOR_RACE_TEST_COUNT; ++test) {
-  do {
+    // Setup sync point dependency to reproduce the race condition of
-    rocksdb::SyncPoint::GetInstance()->ClearTrace();
+    // a log file moved to archived dir, in the middle of GetSortedWalFiles
-    rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+    rocksdb::SyncPoint::GetInstance()->LoadDependency(
-    Options options = OptionsForLogIterTest();
+      { { sync_points[test][0], sync_points[test][1] },
-    DestroyAndReopen(&options);
+        { sync_points[test][2], sync_points[test][3] },
-    Put("key1", DummyString(1024));
+      });
-    dbfull()->Flush(FlushOptions());
+
-    Put("key2", DummyString(1024));
+    do {
-    dbfull()->Flush(FlushOptions());
+      rocksdb::SyncPoint::GetInstance()->ClearTrace();
-    Put("key3", DummyString(1024));
+      rocksdb::SyncPoint::GetInstance()->DisableProcessing();
-    dbfull()->Flush(FlushOptions());
+      Options options = OptionsForLogIterTest();
-    Put("key4", DummyString(1024));
+      DestroyAndReopen(&options);
-    ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), 4U);
+      Put("key1", DummyString(1024));
-
+      dbfull()->Flush(FlushOptions());
-    {
+      Put("key2", DummyString(1024));
-      auto iter = OpenTransactionLogIter(0);
+      dbfull()->Flush(FlushOptions());
-      ExpectRecords(4, iter);
+      Put("key3", DummyString(1024));
-    }
+      dbfull()->Flush(FlushOptions());
-
+      Put("key4", DummyString(1024));
-    rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+      ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), 4U);
-    // trigger async flush, and log move. Well, log move will
+
-    // wait until the GetSortedWalFiles:1 to reproduce the race
+      {
-    // condition
+        auto iter = OpenTransactionLogIter(0);
-    FlushOptions flush_options;
+        ExpectRecords(4, iter);
-    flush_options.wait = false;
+      }
-    dbfull()->Flush(flush_options);
 
-    // "key5" would be written in a new memtable and log
+      rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-    Put("key5", DummyString(1024));
+      // trigger async flush, and log move. Well, log move will
-    {
+      // wait until the GetSortedWalFiles:1 to reproduce the race
-      // this iter would miss "key4" if not fixed
+      // condition
-      auto iter = OpenTransactionLogIter(0);
+      FlushOptions flush_options;
-      ExpectRecords(5, iter);
+      flush_options.wait = false;
-    }
+      dbfull()->Flush(flush_options);
-  } while (ChangeCompactOptions());
+
+      // "key5" would be written in a new memtable and log
+      Put("key5", DummyString(1024));
+      {
+        // this iter would miss "key4" if not fixed
+        auto iter = OpenTransactionLogIter(0);
+        ExpectRecords(5, iter);
+      }
+    } while (ChangeCompactOptions());
+  }
 }
 #endif
 

db/transaction_log_impl.cc

@@ -93,6 +93,7 @@ void TransactionLogIteratorImpl::SeekToStartSequence(
   Status s = OpenLogReader(files_->at(startFileIndex).get());
   if (!s.ok()) {
     currentStatus_ = s;
+    reporter_.Info(currentStatus_.ToString().c_str());
     return;
   }
   while (RestrictedRead(&record, &scratch)) {