fix deadlock with enable_pipelined_write=true and max_successive_merges > 0
Summary: fix this https://github.com/facebook/rocksdb/issues/3916 Closes https://github.com/facebook/rocksdb/pull/3923 Differential Revision: D8215192 Pulled By: yiwu-arbug fbshipit-source-id: a4c2f839a91d92dc70906d2b7c6de0fe014a2422
This commit is contained in:
parent
aaac6cd16f
commit
a35451eaa4
@ -7,6 +7,9 @@
|
||||
### New Features
|
||||
* Changes the format of index blocks by storing the key in their raw form rather than converting them to InternalKey. This saves 8 bytes per index key. The feature is backward compatbile but not forward compatible. It is disabled by default unless format_version 3 or above is used.
|
||||
|
||||
### Bug Fixes
|
||||
* fix deadlock with enable_pipelined_write=true and max_successive_merges > 0
|
||||
|
||||
## 5.14.0 (5/16/2018)
|
||||
### Public API Change
|
||||
* Add a BlockBasedTableOption to align uncompressed data blocks on the smaller of block size or page size boundary, to reduce flash reads by avoiding reads spanning 4K pages.
|
||||
|
@ -1281,7 +1281,11 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context,
|
||||
// In case of pipelined write is enabled, wait for all pending memtable
|
||||
// writers.
|
||||
if (immutable_db_options_.enable_pipelined_write) {
|
||||
// Memtable writers may call DB::Get in case max_successive_merges > 0,
|
||||
// which may lock mutex. Unlocking mutex here to avoid deadlock.
|
||||
mutex_.Unlock();
|
||||
write_thread_.WaitForMemTableWriters();
|
||||
mutex_.Lock();
|
||||
}
|
||||
|
||||
// Attempt to switch to a new memtable and trigger flush of old.
|
||||
|
Loading…
Reference in New Issue
Block a user