Fix bug in auto_readahead_size that returned wrong key (facebook#12229)

Summary:
IndexType::kBinarySearchWithFirstKey +
BlockCacheLookupForReadAheadSize enabled => FindNextUserEntryInternal assertion fails or iterator lands at a wrong key because BlockCacheLookupForReadAheadSize moves the index_iter_ and in internal_wrapper.h, result_.key didn't update and pointed to wrong key. Also ikey_ was also pointing to iter_.key() instead of copying the key.

Pull Request resolved: facebook#12229

Test Plan:
```
 rm -rf /dev/shm/rocksdb_test/rocksdb_crashtest_blackbox_alt3 /dev/shm/rocksdb_test/rocksdb_crashtest_expected_alt3
mkdir /dev/shm/rocksdb_test/rocksdb_crashtest_blackbox_alt3 /dev/shm/rocksdb_test/rocksdb_crashtest_expected_alt3
./db_stress -threads=1 --acquire_snapshot_one_in=0 --adaptive_readahead=0 --allow_concurrent_memtable_write=0 --allow_data_in_errors=True --allow_setting_blob_options_dynamically=0 --async_io=0 --auto_readahead_size=1 --avoid_flush_during_recovery=0 --avoid_unnecessary_blocking_io=1 --backup_max_size=0 --backup_one_in=0 --batch_protection_bytes_per_key=0 --blob_cache_size=0 --blob_compaction_readahead_size=0 --blob_compression_type=lz4 --blob_file_size=0 --blob_file_starting_level=0 --blob_garbage_collection_age_cutoff=0 --blob_garbage_collection_force_threshold=0 --block_protection_bytes_per_key=0 --block_size=2048 --bloom_before_level=2147483646 --bloom_bits=15 --bottommost_compression_type=snappy --bottommost_file_compaction_delay=0 --bytes_per_sync=0 --cache_index_and_filter_blocks=0 --cache_size=8388608 --cache_type=lru_cache --charge_compression_dictionary_building_buffer=0 --charge_file_metadata=0 --charge_filter_construction=0 --charge_table_reader=0 --checkpoint_one_in=0 --checksum_type=kCRC32c --clear_column_family_one_in=0 --column_families=1 --compact_files_one_in=0 --compact_range_one_in=0 --compaction_pri=1 --compaction_readahead_size=0 --compaction_ttl=0 --compressed_secondary_cache_size=0 --compression_checksum=0 --compression_max_dict_buffer_bytes=511 --compression_max_dict_bytes=16384 --compression_parallel_threads=1 --compression_type=none --compression_use_zstd_dict_trainer=1 --compression_zstd_max_train_bytes=0 --continuous_verification_interval=0 --data_block_index_type=1 --db=/dev/shm/rocksdb_test/rocksdb_crashtest_blackbox_alt3 --db_write_buffer_size=0 --delpercent=0 --delrangepercent=0 --destroy_db_initially=0 --detect_filter_construct_corruption=0 --disable_wal=0 --enable_blob_files=0 --enable_blob_garbage_collection=0 --enable_compaction_filter=0 --enable_pipelined_write=0 --enable_thread_tracking=1 --expected_values_dir=/dev/shm/rocksdb_test/rocksdb_crashtest_expected_alt3 --fail_if_options_file_error=1 --fifo_allow_compaction=0 --file_checksum_impl=crc32c --flush_one_in=1000000 --format_version=3 --get_current_wal_file_one_in=0 --get_live_files_one_in=0 --get_property_one_in=0 --get_sorted_wal_files_one_in=0 --index_block_restart_interval=13 --index_type=3 --ingest_external_file_one_in=10 --initial_auto_readahead_size=0 --iterpercent=55 --key_len_percent_dist=1,30,69 --level_compaction_dynamic_level_bytes=0 --lock_wal_one_in=0 --long_running_snapshots=0 --manual_wal_flush_one_in=0 --mark_for_compaction_one_file_in=0 --max_auto_readahead_size=0 --max_background_compactions=1 --max_bytes_for_level_base=67108864 --max_key=100000 --max_key_len=3 --max_manifest_file_size=1073741824 --max_write_batch_group_size_bytes=16 --max_write_buffer_number=10 --max_write_buffer_size_to_maintain=4194304 --memtable_max_range_deletions=1000 --memtable_prefix_bloom_size_ratio=0.5 --memtable_protection_bytes_per_key=0 --memtable_whole_key_filtering=0 --memtablerep=skip_list --min_blob_size=8 --min_write_buffer_number_to_merge=2 --mmap_read=0 --mock_direct_io=False --nooverwritepercent=1 --num_file_reads_for_auto_readahead=2 --open_files=-1 --open_metadata_write_fault_one_in=0 --open_read_fault_one_in=0 --open_write_fault_one_in=0 --ops_per_thread=10000000 --optimize_filters_for_memory=0 --paranoid_file_checks=0 --partition_filters=0 --partition_pinning=0 --pause_background_one_in=0 --periodic_compaction_seconds=0 --prefix_size=1 --prefixpercent=0 --prepopulate_block_cache=0 --preserve_internal_time_seconds=0 --progress_reports=0 --read_fault_one_in=0 --readahead_size=1 --readpercent=45 --recycle_log_file_num=1 --reopen=0 --secondary_cache_fault_one_in=0 --secondary_cache_uri= --set_options_one_in=0 --snapshot_hold_ops=0 --sst_file_manager_bytes_per_sec=0 --sst_file_manager_bytes_per_truncate=0 --stats_dump_period_sec=600 --subcompactions=1 --sync=0 --sync_fault_injection=0 --target_file_size_base=16777216 --target_file_size_multiplier=1 --test_batches_snapshots=0 --top_level_index_pinning=0 --unpartitioned_pinning=0 --use_blob_cache=0 --use_direct_io_for_flush_and_compaction=0 --use_direct_reads=0 --use_full_merge_v1=0 --use_get_entity=0 --use_merge=0 --use_multi_get_entity=0 --use_multiget=0 --use_put_entity_one_in=0 --use_shared_block_and_blob_cache=0 --use_write_buffer_manager=0 --user_timestamp_size=0 --value_size_mult=32 --verification_only=0 --verify_checksum=0 --verify_checksum_one_in=0 --verify_db_one_in=0 --verify_file_checksums_one_in=0 --verify_iterator_with_expected_state_one_in=1 --verify_sst_unique_id_in_manifest=0 --wal_bytes_per_sync=0 --wal_compression=none --write_buffer_size=33554432 --write_dbid_to_manifest=0 --write_fault_one_in=0 --writepercent=0 > repro.out
Verification failed. Expected state has key 0000000000000077000000000000004178, iterator is at key 0000000000000077000000000000008A78
Column family: default, op_logs: S 0000000000000077000000000000003D7878787878 NNNN
No writes or ops?
Verification failed :(
```

Reviewed By: ajkr

Differential Revision: D52710655

Pulled By: akankshamahajan15

fbshipit-source-id: 9d2e684e190fb0832bdce3337bce1c6548cd054d

db/db_iter.cc

@@ -395,9 +395,7 @@ bool DBIter::FindNextUserEntryInternal(bool skipping_saved_key,
  case kTypeValue:
  case kTypeBlobIndex:
  case kTypeWideColumnEntity:
- if (!iter_.PrepareValue()) {
- assert(!iter_.status().ok());
- valid_ = false;
+ if (!PrepareValue()) {
  return false;
  }
  if (timestamp_lb_) {
@@ -428,9 +426,7 @@ bool DBIter::FindNextUserEntryInternal(bool skipping_saved_key,
  return true;
  break;
  case kTypeMerge:
- if (!iter_.PrepareValue()) {
- assert(!iter_.status().ok());
- valid_ = false;
+ if (!PrepareValue()) {
  return false;
  }
  saved_key_.SetUserKey(
@@ -575,8 +571,7 @@ bool DBIter::MergeValuesNewToOld() {
  iter_.Next();
  break;
  }
- if (!iter_.PrepareValue()) {
- valid_ = false;
+ if (!PrepareValue()) {
  return false;
  }
 
@@ -902,8 +897,7 @@ bool DBIter::FindValueForCurrentKey() {
  return FindValueForCurrentKeyUsingSeek();
  }
 
- if (!iter_.PrepareValue()) {
- valid_ = false;
+ if (!PrepareValue()) {
  return false;
  }
 
@@ -1142,8 +1136,7 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
  }
  return true;
  }
- if (!iter_.PrepareValue()) {
- valid_ = false;
+ if (!PrepareValue()) {
  return false;
  }
  if (timestamp_size_ > 0) {
@@ -1208,8 +1201,7 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
  ikey.type == kTypeDeletionWithTimestamp) {
  break;
  }
- if (!iter_.PrepareValue()) {
- valid_ = false;
+ if (!PrepareValue()) {
  return false;
  }
 

db/db_iter.h

@@ -333,6 +333,22 @@ class DBIter final : public Iterator {
  bool MergeWithPlainBaseValue(const Slice& value, const Slice& user_key);
  bool MergeWithWideColumnBaseValue(const Slice& entity, const Slice& user_key);
 
+ bool PrepareValue() {
+ if (!iter_.PrepareValue()) {
+ assert(!iter_.status().ok());
+ valid_ = false;
+ return false;
+ }
+ // ikey_ could change as BlockBasedTableIterator does Block cache
+ // lookup and index_iter_ could point to different block resulting
+ // in ikey_ pointing to wrong key. So ikey_ needs to be updated in
+ // case of Seek/Next calls to point to right key again.
+ if (!ParseKey(&ikey_)) {
+ return false;
+ }
+ return true;
+ }
+
  const SliceTransform* prefix_extractor_;
  Env* const env_;
  SystemClock* clock_;

table/iterator_wrapper.h

@@ -101,6 +101,7 @@ class IteratorWrapperBase {
  }
  if (iter_->PrepareValue()) {
  result_.value_prepared = true;
+ result_.key = iter_->key();
  return true;
  }
 

unreleased_history/bug_fixes/auto_readahead_size.md

@@ -0,0 +1 @@
+Fix bug in auto_readahead_size that combined with IndexType::kBinarySearchWithFirstKey + fails or iterator lands at a wrong key