titan
titan copied to clipboard
tikv
A RocksDB plugin for key-value separation, inspired by WiscKey.
The issue is here #238. I push 3 commits. One is add a status check. One is change the `void BatchWriteNewIndices(BlobFileBuilder::OutContexts& contexts, Status* s)` to `Status BatchWriteNewIndices(BlobFileBuilder::OutContexts& contexts)` One is...
# Description Titan now can use delta compression. Here is my code [repository](https://github.com/apple-ouyang/titan/tree/1io) Acording to the test result, the compression ratio for compressed record can reach up to 77.88x! However...
In BlobGCJob::InstallOutputBlobFiles, which is blob_gc_joob.cc:415, the status are not checked. ```C++ for (auto& builder : blob_file_builders_) { BlobFileBuilder::OutContexts contexts; s = builder.second->Finish(&contexts); // Status not check BatchWriteNewIndices(contexts, &s); // Status...
In BlobGCJob::DoRunGC(), the `file_size` is always = 0. I guess this variable is used to accumulate the file size of blob file. Once the `file_size >= blob_gc_->titan_cf_options().blob_file_target_size`, Titan will create...
The test version is the latested master(2ad9c586345b606718362351f5c88e3c8348c4f4) After cmake and compiling, run the: ```bash ctest -R titan ``` the only test that not pass is the titan_titan_db_test. Here is the...
Signed-off-by: jason 我们线上使用titan的时候经常发现get会产生几百毫秒延时，原因是get命令创建snapshot的时候会mutex_.Lock()。但BlobGCJob对象在析构的时候会持有mutex_，该对象中保存了需要回写sst中的kv，如果需要回写的kv很多（线下测试可能会有几十万条），那么该对象会很大，析构的时候比较耗时，持有mutex_的时间就会比较长，导致get命令加锁被阻塞。所以建议析构大对象的时候，不要持有锁。我们修复后，线上已经跑了一个月，get已经不会产生延时抖动。 优化前：一天get超过200ms延时大概在10w条左右 优化后：一天get超过200ms延时大概在100条左右
 Value of metric `live-blob-size` is much larger than total size of blob file, and when most of data has been removed, it is still so larger (ten times than...
Limit GC and Titan read IO with rate_limiter, to avoid burst of IO starve online read and write.
