Extend Get/MultiGet deadline support to table open (facebook#6982)

Summary:
Current implementation of the ```read_options.deadline``` option only checks the deadline for random file reads during point lookups. This PR extends the checks to file opens, prefetches and preloads as part of table open.

The main changes are in the ```BlockBasedTable```, partitioned index and filter readers, and ```TableCache``` to take ReadOptions as an additional parameter. In ```BlockBasedTable::Open```, in order to retain existing behavior w.r.t checksum verification and block cache usage, we filter out most of the options in ```ReadOptions``` except ```deadline```. However, having the ```ReadOptions``` gives us more flexibility to honor other options like verify_checksums, fill_cache etc. in the future.

Additional changes in callsites due to function signature changes in ```NewTableReader()``` and ```FilePrefetchBuffer```.
Pull Request resolved: facebook#6982

Test Plan: Add new unit tests in db_basic_test

Reviewed By: riversand963

Differential Revision: D22219515

Pulled By: anand1976

fbshipit-source-id: 8a3b92f4a889808013838603aa3ca35229cd501b

db/db_impl/db_impl.cc

@@ -2685,8 +2685,6 @@ Iterator* DBImpl::NewIterator(const ReadOptions& read_options,
     return NewErrorIterator(
         Status::NotSupported("Managed iterator is not supported anymore."));
   }
-  // We will eventually support deadline for iterators too, but safeguard
-  // for now
   if (read_options.deadline != std::chrono::microseconds::zero()) {
     return NewErrorIterator(
         Status::NotSupported("ReadOptions deadline is not supported"));

db/plain_table_db_test.cc

@@ -336,8 +336,9 @@ class TestPlainTableFactory : public PlainTableFactory {
         column_family_id_(column_family_id),
         column_family_name_(std::move(column_family_name)) {}
 
+  using PlainTableFactory::NewTableReader;
   Status NewTableReader(
-      const TableReaderOptions& table_reader_options,
+      const ReadOptions& /*ro*/, const TableReaderOptions& table_reader_options,
       std::unique_ptr<RandomAccessFileReader>&& file, uint64_t file_size,
       std::unique_ptr<TableReader>* table,
       bool /*prefetch_index_and_filter_in_cache*/) const override {

db/table_cache.cc

@@ -13,6 +13,7 @@
 #include "db/range_tombstone_fragmenter.h"
 #include "db/snapshot_impl.h"
 #include "db/version_edit.h"
+#include "file/file_util.h"
 #include "file/filename.h"
 #include "file/random_access_file_reader.h"
 #include "monitoring/perf_context_imp.h"
@@ -92,7 +93,7 @@ void TableCache::ReleaseHandle(Cache::Handle* handle) {
 }
 
 Status TableCache::GetTableReader(
-    const FileOptions& file_options,
+    const ReadOptions& ro, const FileOptions& file_options,
     const InternalKeyComparator& internal_comparator, const FileDescriptor& fd,
     bool sequential_mode, bool record_read_stats, HistogramImpl* file_read_hist,
     std::unique_ptr<TableReader>* table_reader,
@@ -102,12 +103,19 @@ Status TableCache::GetTableReader(
   std::string fname =
       TableFileName(ioptions_.cf_paths, fd.GetNumber(), fd.GetPathId());
   std::unique_ptr<FSRandomAccessFile> file;
-  Status s = ioptions_.fs->NewRandomAccessFile(fname, file_options, &file,
-                                               nullptr);
+  FileOptions fopts = file_options;
+  Status s = PrepareIOFromReadOptions(ro, ioptions_.env, fopts.io_options);
+  if (s.ok()) {
+    s = ioptions_.fs->NewRandomAccessFile(fname, fopts, &file, nullptr);
+  }
   RecordTick(ioptions_.statistics, NO_FILE_OPENS);
   if (s.IsPathNotFound()) {
     fname = Rocks2LevelTableFileName(fname);
-    s = ioptions_.fs->NewRandomAccessFile(fname, file_options, &file, nullptr);
+    s = PrepareIOFromReadOptions(ro, ioptions_.env, fopts.io_options);
+    if (s.ok()) {
+      s = ioptions_.fs->NewRandomAccessFile(fname, file_options, &file,
+                                            nullptr);
+    }
     RecordTick(ioptions_.statistics, NO_FILE_OPENS);
   }
 
@@ -122,6 +130,7 @@ Status TableCache::GetTableReader(
             record_read_stats ? ioptions_.statistics : nullptr, SST_READ_MICROS,
             file_read_hist, ioptions_.rate_limiter, ioptions_.listeners));
     s = ioptions_.table_factory->NewTableReader(
+        ro,
         TableReaderOptions(ioptions_, prefix_extractor, file_options,
                            internal_comparator, skip_filters, immortal_tables_,
                            false /* force_direct_prefetch */, level,
@@ -141,7 +150,8 @@ void TableCache::EraseHandle(const FileDescriptor& fd, Cache::Handle* handle) {
   cache_->Erase(key);
 }
 
-Status TableCache::FindTable(const FileOptions& file_options,
+Status TableCache::FindTable(const ReadOptions& ro,
+                             const FileOptions& file_options,
                              const InternalKeyComparator& internal_comparator,
                              const FileDescriptor& fd, Cache::Handle** handle,
                              const SliceTransform* prefix_extractor,
@@ -169,7 +179,7 @@ Status TableCache::FindTable(const FileOptions& file_options,
     }
 
     std::unique_ptr<TableReader> table_reader;
-    s = GetTableReader(file_options, internal_comparator, fd,
+    s = GetTableReader(ro, file_options, internal_comparator, fd,
                        false /* sequential mode */, record_read_stats,
                        file_read_hist, &table_reader, prefix_extractor,
                        skip_filters, level, prefetch_index_and_filter_in_cache,
@@ -212,12 +222,12 @@ InternalIterator* TableCache::NewIterator(
   auto& fd = file_meta.fd;
   table_reader = fd.table_reader;
   if (table_reader == nullptr) {
-    s = FindTable(file_options, icomparator, fd, &handle, prefix_extractor,
-                  options.read_tier == kBlockCacheTier /* no_io */,
-                  !for_compaction /* record_read_stats */, file_read_hist,
-                  skip_filters, level,
-                  true /* prefetch_index_and_filter_in_cache */,
-                  max_file_size_for_l0_meta_pin);
+    s = FindTable(
+        options, file_options, icomparator, fd, &handle, prefix_extractor,
+        options.read_tier == kBlockCacheTier /* no_io */,
+        !for_compaction /* record_read_stats */, file_read_hist, skip_filters,
+        level, true /* prefetch_index_and_filter_in_cache */,
+        max_file_size_for_l0_meta_pin);
     if (s.ok()) {
       table_reader = GetTableReaderFromHandle(handle);
     }
@@ -288,7 +298,7 @@ Status TableCache::GetRangeTombstoneIterator(
   TableReader* t = fd.table_reader;
   Cache::Handle* handle = nullptr;
   if (t == nullptr) {
-    s = FindTable(file_options_, internal_comparator, fd, &handle);
+    s = FindTable(options, file_options_, internal_comparator, fd, &handle);
     if (s.ok()) {
       t = GetTableReaderFromHandle(handle);
     }
@@ -403,7 +413,7 @@ Status TableCache::Get(const ReadOptions& options,
   Cache::Handle* handle = nullptr;
   if (!done && s.ok()) {
     if (t == nullptr) {
-      s = FindTable(file_options_, internal_comparator, fd, &handle,
+      s = FindTable(options, file_options_, internal_comparator, fd, &handle,
                     prefix_extractor,
                     options.read_tier == kBlockCacheTier /* no_io */,
                     true /* record_read_stats */, file_read_hist, skip_filters,
@@ -506,8 +516,8 @@ Status TableCache::MultiGet(const ReadOptions& options,
   if (s.ok() && !table_range.empty()) {
     if (t == nullptr) {
       s = FindTable(
-          file_options_, internal_comparator, fd, &handle, prefix_extractor,
-          options.read_tier == kBlockCacheTier /* no_io */,
+          options, file_options_, internal_comparator, fd, &handle,
+          prefix_extractor, options.read_tier == kBlockCacheTier /* no_io */,
           true /* record_read_stats */, file_read_hist, skip_filters, level);
       TEST_SYNC_POINT_CALLBACK("TableCache::MultiGet:FindTable", &s);
       if (s.ok()) {
@@ -591,8 +601,8 @@ Status TableCache::GetTableProperties(
   }
 
   Cache::Handle* table_handle = nullptr;
-  s = FindTable(file_options, internal_comparator, fd, &table_handle,
-                prefix_extractor, no_io);
+  s = FindTable(ReadOptions(), file_options, internal_comparator, fd,
+                &table_handle, prefix_extractor, no_io);
   if (!s.ok()) {
     return s;
   }
@@ -615,8 +625,8 @@ size_t TableCache::GetMemoryUsageByTableReader(
   }
 
   Cache::Handle* table_handle = nullptr;
-  s = FindTable(file_options, internal_comparator, fd, &table_handle,
-                prefix_extractor, true);
+  s = FindTable(ReadOptions(), file_options, internal_comparator, fd,
+                &table_handle, prefix_extractor, true);
   if (!s.ok()) {
     return 0;
   }
@@ -640,8 +650,8 @@ uint64_t TableCache::ApproximateOffsetOf(
   Cache::Handle* table_handle = nullptr;
   if (table_reader == nullptr) {
     const bool for_compaction = (caller == TableReaderCaller::kCompaction);
-    Status s = FindTable(file_options_, internal_comparator, fd, &table_handle,
-                         prefix_extractor, false /* no_io */,
+    Status s = FindTable(ReadOptions(), file_options_, internal_comparator, fd,
+                         &table_handle, prefix_extractor, false /* no_io */,
                          !for_compaction /* record_read_stats */);
     if (s.ok()) {
       table_reader = GetTableReaderFromHandle(table_handle);
@@ -667,8 +677,8 @@ uint64_t TableCache::ApproximateSize(
   Cache::Handle* table_handle = nullptr;
   if (table_reader == nullptr) {
     const bool for_compaction = (caller == TableReaderCaller::kCompaction);
-    Status s = FindTable(file_options_, internal_comparator, fd, &table_handle,
-                         prefix_extractor, false /* no_io */,
+    Status s = FindTable(ReadOptions(), file_options_, internal_comparator, fd,
+                         &table_handle, prefix_extractor, false /* no_io */,
                          !for_compaction /* record_read_stats */);
     if (s.ok()) {
       table_reader = GetTableReaderFromHandle(table_handle);

db/table_cache.h

@@ -129,7 +129,7 @@ class TableCache {
   // Find table reader
   // @param skip_filters Disables loading/accessing the filter block
   // @param level == -1 means not specified
-  Status FindTable(const FileOptions& toptions,
+  Status FindTable(const ReadOptions& ro, const FileOptions& toptions,
                    const InternalKeyComparator& internal_comparator,
                    const FileDescriptor& file_fd, Cache::Handle**,
                    const SliceTransform* prefix_extractor = nullptr,
@@ -195,7 +195,7 @@ class TableCache {
 
  private:
   // Build a table reader
-  Status GetTableReader(const FileOptions& file_options,
+  Status GetTableReader(const ReadOptions& ro, const FileOptions& file_options,
                         const InternalKeyComparator& internal_comparator,
                         const FileDescriptor& fd, bool sequential_mode,
                         bool record_read_stats, HistogramImpl* file_read_hist,

db/version_builder.cc

@@ -931,9 +931,10 @@ class VersionBuilder::Rep {
         auto* file_meta = files_meta[file_idx].first;
         int level = files_meta[file_idx].second;
         statuses[file_idx] = table_cache_->FindTable(
-            file_options_, *(base_vstorage_->InternalComparator()),
-            file_meta->fd, &file_meta->table_reader_handle, prefix_extractor,
-            false /*no_io */, true /* record_read_stats */,
+            ReadOptions(), file_options_,
+            *(base_vstorage_->InternalComparator()), file_meta->fd,
+            &file_meta->table_reader_handle, prefix_extractor, false /*no_io */,
+            true /* record_read_stats */,
             internal_stats->GetFileReadHist(level), false, level,
             prefetch_index_and_filter_in_cache, max_file_size_for_l0_meta_pin);
         if (file_meta->table_reader_handle != nullptr) {

file/file_prefetch_buffer.cc

@@ -21,7 +21,8 @@
 #include "util/rate_limiter.h"
 
 namespace ROCKSDB_NAMESPACE {
-Status FilePrefetchBuffer::Prefetch(RandomAccessFileReader* reader,
+Status FilePrefetchBuffer::Prefetch(const IOOptions& opts,
+                                    RandomAccessFileReader* reader,
                                     uint64_t offset, size_t n,
                                     bool for_compaction) {
   if (!enable_ || reader == nullptr) {
@@ -87,7 +88,7 @@ Status FilePrefetchBuffer::Prefetch(RandomAccessFileReader* reader,
 
   Slice result;
   size_t read_len = static_cast<size_t>(roundup_len - chunk_len);
-  s = reader->Read(IOOptions(), rounddown_offset + chunk_len, read_len, &result,
+  s = reader->Read(opts, rounddown_offset + chunk_len, read_len, &result,
                    buffer_.BufferStart() + chunk_len, nullptr, for_compaction);
 #ifndef NDEBUG
   if (!s.ok() || result.size() < read_len) {
@@ -103,7 +104,8 @@ Status FilePrefetchBuffer::Prefetch(RandomAccessFileReader* reader,
   return s;
 }
 
-bool FilePrefetchBuffer::TryReadFromCache(uint64_t offset, size_t n,
+bool FilePrefetchBuffer::TryReadFromCache(const IOOptions& opts,
+                                          uint64_t offset, size_t n,
                                           Slice* result, bool for_compaction) {
   if (track_min_offset_ && offset < min_offset_read_) {
     min_offset_read_ = static_cast<size_t>(offset);
@@ -122,10 +124,11 @@ bool FilePrefetchBuffer::TryReadFromCache(uint64_t offset, size_t n,
       assert(max_readahead_size_ >= readahead_size_);
       Status s;
       if (for_compaction) {
-        s = Prefetch(file_reader_, offset, std::max(n, readahead_size_),
+        s = Prefetch(opts, file_reader_, offset, std::max(n, readahead_size_),
                      for_compaction);
       } else {
-        s = Prefetch(file_reader_, offset, n + readahead_size_, for_compaction);
+        s = Prefetch(opts, file_reader_, offset, n + readahead_size_,
+                     for_compaction);
       }
       if (!s.ok()) {
         return false;

file/file_prefetch_buffer.h

@@ -11,9 +11,11 @@
 #include <atomic>
 #include <sstream>
 #include <string>
+
 #include "file/random_access_file_reader.h"
 #include "port/port.h"
 #include "rocksdb/env.h"
+#include "rocksdb/options.h"
 #include "util/aligned_buffer.h"
 
 namespace ROCKSDB_NAMESPACE {
@@ -59,8 +61,8 @@ class FilePrefetchBuffer {
   // offset : the file offset to start reading from.
   // n      : the number of bytes to read.
   // for_compaction : if prefetch is done for compaction read.
-  Status Prefetch(RandomAccessFileReader* reader, uint64_t offset, size_t n,
-                  bool for_compaction = false);
+  Status Prefetch(const IOOptions& opts, RandomAccessFileReader* reader,
+                  uint64_t offset, size_t n, bool for_compaction = false);
 
   // Tries returning the data for a file raed from this buffer, if that data is
   // in the buffer.
@@ -72,8 +74,8 @@ class FilePrefetchBuffer {
   // n      : the number of bytes.
   // result : output buffer to put the data into.
   // for_compaction : if cache read is done for compaction read.
-  bool TryReadFromCache(uint64_t offset, size_t n, Slice* result,
-                        bool for_compaction = false);
+  bool TryReadFromCache(const IOOptions& opts, uint64_t offset, size_t n,
+                        Slice* result, bool for_compaction = false);
 
   // The minimum `offset` ever passed to TryReadFromCache(). This will nly be
   // tracked if track_min_offset = true.

file/file_util.cc

@@ -166,10 +166,11 @@ IOStatus GenerateOneFileChecksum(FileSystem* fs, const std::string& file_path,
 
   Slice slice;
   uint64_t offset = 0;
+  IOOptions opts;
   while (size > 0) {
     size_t bytes_to_read =
         static_cast<size_t>(std::min(uint64_t{readahead_size}, size));
-    if (!prefetch_buffer.TryReadFromCache(offset, bytes_to_read, &slice,
+    if (!prefetch_buffer.TryReadFromCache(opts, offset, bytes_to_read, &slice,
                                           false)) {
       return IOStatus::Corruption("file read failed");
     }

file/random_access_file_reader.h

@@ -11,10 +11,12 @@
 #include <atomic>
 #include <sstream>
 #include <string>
+
 #include "port/port.h"
 #include "rocksdb/env.h"
 #include "rocksdb/file_system.h"
 #include "rocksdb/listener.h"
+#include "rocksdb/options.h"
 #include "rocksdb/rate_limiter.h"
 #include "util/aligned_buffer.h"
 

include/rocksdb/table.h

@@ -556,7 +556,19 @@ class TableFactory {
       const TableReaderOptions& table_reader_options,
       std::unique_ptr<RandomAccessFileReader>&& file, uint64_t file_size,
       std::unique_ptr<TableReader>* table_reader,
-      bool prefetch_index_and_filter_in_cache = true) const = 0;
+      bool prefetch_index_and_filter_in_cache = true) const {
+    ReadOptions ro;
+    return NewTableReader(ro, table_reader_options, std::move(file), file_size,
+                          table_reader, prefetch_index_and_filter_in_cache);
+  }
+
+  // Overload of the above function that allows the caller to pass in a
+  // ReadOptions
+  virtual Status NewTableReader(
+      const ReadOptions& ro, const TableReaderOptions& table_reader_options,
+      std::unique_ptr<RandomAccessFileReader>&& file, uint64_t file_size,
+      std::unique_ptr<TableReader>* table_reader,
+      bool prefetch_index_and_filter_in_cache) const = 0;
 
   // Return a table builder to write to a file for this table type.
   //

table/adaptive/adaptive_table_factory.cc

@@ -42,12 +42,13 @@ extern const uint64_t kLegacyBlockBasedTableMagicNumber;
 extern const uint64_t kCuckooTableMagicNumber;
 
 Status AdaptiveTableFactory::NewTableReader(
-    const TableReaderOptions& table_reader_options,
+    const ReadOptions& ro, const TableReaderOptions& table_reader_options,
     std::unique_ptr<RandomAccessFileReader>&& file, uint64_t file_size,
     std::unique_ptr<TableReader>* table,
-    bool /*prefetch_index_and_filter_in_cache*/) const {
+    bool prefetch_index_and_filter_in_cache) const {
   Footer footer;
-  auto s = ReadFooterFromFile(file.get(), nullptr /* prefetch_buffer */,
+  IOOptions opts;
+  auto s = ReadFooterFromFile(opts, file.get(), nullptr /* prefetch_buffer */,
                               file_size, &footer);
   if (!s.ok()) {
     return s;
@@ -59,7 +60,8 @@ Status AdaptiveTableFactory::NewTableReader(
   } else if (footer.table_magic_number() == kBlockBasedTableMagicNumber ||
       footer.table_magic_number() == kLegacyBlockBasedTableMagicNumber) {
     return block_based_table_factory_->NewTableReader(
-        table_reader_options, std::move(file), file_size, table);
+        ro, table_reader_options, std::move(file), file_size, table,
+        prefetch_index_and_filter_in_cache);
   } else if (footer.table_magic_number() == kCuckooTableMagicNumber) {
     return cuckoo_table_factory_->NewTableReader(
         table_reader_options, std::move(file), file_size, table);

table/adaptive/adaptive_table_factory.h

@@ -33,8 +33,9 @@ class AdaptiveTableFactory : public TableFactory {
 
   const char* Name() const override { return "AdaptiveTableFactory"; }
 
+  using TableFactory::NewTableReader;
   Status NewTableReader(
-      const TableReaderOptions& table_reader_options,
+      const ReadOptions& ro, const TableReaderOptions& table_reader_options,
       std::unique_ptr<RandomAccessFileReader>&& file, uint64_t file_size,
       std::unique_ptr<TableReader>* table,
       bool prefetch_index_and_filter_in_cache = true) const override;

table/block_based/binary_search_index_reader.cc

@@ -10,9 +10,9 @@
 
 namespace ROCKSDB_NAMESPACE {
 Status BinarySearchIndexReader::Create(
-    const BlockBasedTable* table, FilePrefetchBuffer* prefetch_buffer,
-    bool use_cache, bool prefetch, bool pin,
-    BlockCacheLookupContext* lookup_context,
+    const BlockBasedTable* table, const ReadOptions& ro,
+    FilePrefetchBuffer* prefetch_buffer, bool use_cache, bool prefetch,
+    bool pin, BlockCacheLookupContext* lookup_context,
     std::unique_ptr<IndexReader>* index_reader) {
   assert(table != nullptr);
   assert(table->get_rep());
@@ -22,7 +22,7 @@ Status BinarySearchIndexReader::Create(
   CachableEntry<Block> index_block;
   if (prefetch || !use_cache) {
     const Status s =
-        ReadIndexBlock(table, prefetch_buffer, ReadOptions(), use_cache,
+        ReadIndexBlock(table, prefetch_buffer, ro, use_cache,
                        /*get_context=*/nullptr, lookup_context, &index_block);
     if (!s.ok()) {
       return s;