Enable checkpoint of read-only db (#4681)

Summary:
1. DBImplReadOnly::GetLiveFiles should not return NotSupported. Instead, it
   should call DBImpl::GetLiveFiles(flush_memtable=false).
2. In DBImp::Recover, we should also recover the OPTIONS file name and/or
   number so that an immediate subsequent GetLiveFiles will get the correct
   OPTIONS name.
Pull Request resolved: facebook/rocksdb#4681

Differential Revision: D13069205

Pulled By: riversand963

fbshipit-source-id: 3e6a0174307d06db5a01feb099b306cea1f7f88a

HISTORY.md

@@ -1,12 +1,4 @@
 # Rocksdb Change Log
-## Unreleased
-### New Features
-
-### Public API Change
-
-### Bug Fixes
-* Fix a deadlock caused by compaction and file ingestion waiting for each other in the event of write stalls.
-
 ## 5.18.0 (11/30/2018)
 ### New Features
 * Introduced `JemallocNodumpAllocator` memory allocator. When being use, block cache will be excluded from core dump.
@@ -18,6 +10,7 @@
 * Add xxhash64 checksum support
 * Introduced `MemoryAllocator`, which lets the user specify custom memory allocator for block based table.
 * Improved `DeleteRange` to prevent read performance degradation. The feature is no longer marked as experimental.
+* Enabled checkpoint on readonly db (DBImplReadOnly).
 
 ### Public API Change
 * `DBOptions::use_direct_reads` now affects reads issued by `BackupEngine` on the database's SSTs.
@@ -34,6 +27,7 @@
 * Fixed Get correctness bug in the presence of range tombstones where merge operands covered by a range tombstone always result in NotFound.
 * Start populating `NO_FILE_CLOSES` ticker statistic, which was always zero previously.
 * The default value of NewBloomFilterPolicy()'s argument use_block_based_builder is changed to false. Note that this new default may cause large temp memory usage when building very large SST files.
+* Fix a deadlock caused by compaction and file ingestion waiting for each other in the event of write stalls.
 
 ## 5.17.0 (10/05/2018)
 ### Public API Change

db/compacted_db_impl.h

@@ -67,10 +67,11 @@ class CompactedDBImpl : public DBImpl {
   virtual Status EnableFileDeletions(bool /*force*/) override {
     return Status::NotSupported("Not supported in compacted db mode.");
   }
-  virtual Status GetLiveFiles(std::vector<std::string>&,
-                              uint64_t* /*manifest_file_size*/,
-                              bool /*flush_memtable*/ = true) override {
-    return Status::NotSupported("Not supported in compacted db mode.");
+  virtual Status GetLiveFiles(std::vector<std::string>& ret,
+                              uint64_t* manifest_file_size,
+                              bool /*flush_memtable*/) override {
+    return DBImpl::GetLiveFiles(ret, manifest_file_size,
+                                false /* flush_memtable */);
   }
   using DBImpl::Flush;
   virtual Status Flush(const FlushOptions& /*options*/,

db/db_impl_open.cc

@@ -474,6 +474,28 @@ Status DBImpl::Recover(
     }
   }
 
+  if (read_only) {
+    // If we are opening as read-only, we need to update options_file_number_
+    // to reflect the most recent OPTIONS file. It does not matter for regular
+    // read-write db instance because options_file_number_ will later be
+    // updated to versions_->NewFileNumber() in RenameTempFileToOptionsFile.
+    std::vector<std::string> file_names;
+    if (s.ok()) {
+      s = env_->GetChildren(GetName(), &file_names);
+    }
+    if (s.ok()) {
+      uint64_t number = 0;
+      uint64_t options_file_number = 0;
+      FileType type;
+      for (const auto& fname : file_names) {
+        if (ParseFileName(fname, &number, &type) && type == kOptionsFile) {
+          options_file_number = std::max(number, options_file_number);
+        }
+      }
+      versions_->options_file_number_ = options_file_number;
+    }
+  }
+
   return s;
 }
 

db/db_impl_readonly.h

@@ -89,10 +89,11 @@ class DBImplReadOnly : public DBImpl {
   virtual Status EnableFileDeletions(bool /*force*/) override {
     return Status::NotSupported("Not supported operation in read only mode.");
   }
-  virtual Status GetLiveFiles(std::vector<std::string>&,
-                              uint64_t* /*manifest_file_size*/,
-                              bool /*flush_memtable*/ = true) override {
-    return Status::NotSupported("Not supported operation in read only mode.");
+  virtual Status GetLiveFiles(std::vector<std::string>& ret,
+                              uint64_t* manifest_file_size,
+                              bool /*flush_memtable*/) override {
+    return DBImpl::GetLiveFiles(ret, manifest_file_size,
+                                false /* flush_memtable */);
   }
 
   using DBImpl::Flush;

utilities/checkpoint/checkpoint_test.cc

@@ -164,6 +164,16 @@ class CheckpointTest : public testing::Test {
     return DB::OpenForReadOnly(options, dbname_, &db_);
   }
 
+  Status ReadOnlyReopenWithColumnFamilies(const std::vector<std::string>& cfs,
+                                          const Options& options) {
+    std::vector<ColumnFamilyDescriptor> column_families;
+    for (const auto& cf : cfs) {
+      column_families.emplace_back(cf, options);
+    }
+    return DB::OpenForReadOnly(options, dbname_, column_families, &handles_,
+                               &db_);
+  }
+
   Status TryReopen(const Options& options) {
     Close();
     last_options_ = options;
@@ -612,6 +622,69 @@ TEST_F(CheckpointTest, CheckpointWithUnsyncedDataDropped) {
   db_ = nullptr;
 }
 
+TEST_F(CheckpointTest, CheckpointReadOnlyDB) {
+  ASSERT_OK(Put("foo", "foo_value"));
+  ASSERT_OK(Flush());
+  Close();
+  Options options = CurrentOptions();
+  ASSERT_OK(ReadOnlyReopen(options));
+  Checkpoint* checkpoint = nullptr;
+  ASSERT_OK(Checkpoint::Create(db_, &checkpoint));
+  ASSERT_OK(checkpoint->CreateCheckpoint(snapshot_name_));
+  delete checkpoint;
+  checkpoint = nullptr;
+  Close();
+  DB* snapshot_db = nullptr;
+  ASSERT_OK(DB::Open(options, snapshot_name_, &snapshot_db));
+  ReadOptions read_opts;
+  std::string get_result;
+  ASSERT_OK(snapshot_db->Get(read_opts, "foo", &get_result));
+  ASSERT_EQ("foo_value", get_result);
+  delete snapshot_db;
+}
+
+TEST_F(CheckpointTest, CheckpointReadOnlyDBWithMultipleColumnFamilies) {
+  Options options = CurrentOptions();
+  CreateAndReopenWithCF({"pikachu", "eevee"}, options);
+  for (int i = 0; i != 3; ++i) {
+    ASSERT_OK(Put(i, "foo", "foo_value"));
+    ASSERT_OK(Flush(i));
+  }
+  Close();
+  Status s = ReadOnlyReopenWithColumnFamilies(
+      {kDefaultColumnFamilyName, "pikachu", "eevee"}, options);
+  ASSERT_OK(s);
+  Checkpoint* checkpoint = nullptr;
+  ASSERT_OK(Checkpoint::Create(db_, &checkpoint));
+  ASSERT_OK(checkpoint->CreateCheckpoint(snapshot_name_));
+  delete checkpoint;
+  checkpoint = nullptr;
+  Close();
+
+  std::vector<ColumnFamilyDescriptor> column_families{
+      {kDefaultColumnFamilyName, options},
+      {"pikachu", options},
+      {"eevee", options}};
+  DB* snapshot_db = nullptr;
+  std::vector<ColumnFamilyHandle*> snapshot_handles;
+  s = DB::Open(options, snapshot_name_, column_families, &snapshot_handles,
+               &snapshot_db);
+  ASSERT_OK(s);
+  ReadOptions read_opts;
+  for (int i = 0; i != 3; ++i) {
+    std::string get_result;
+    s = snapshot_db->Get(read_opts, snapshot_handles[i], "foo", &get_result);
+    ASSERT_OK(s);
+    ASSERT_EQ("foo_value", get_result);
+  }
+
+  for (auto snapshot_h : snapshot_handles) {
+    delete snapshot_h;
+  }
+  snapshot_handles.clear();
+  delete snapshot_db;
+}
+
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {