Mark Callaghan
Mark Callaghan
If you are still pulling changes for Linkbench activity has moved to https://github.com/mdcallag/linkbench because the upstream version was archived at https://github.com/facebookarchive/linkbench On Wed, Mar 22, 2017 at 7:06 PM, Daniel...
Are you open to another PR to fix this? Because I get extremely misleading flamegraphs today with: ``` perf record -g -p $pid -- sleep $perf_secs perf script > o.scr...
This change seems like it can have a big impact but the commit message content is too brief. Please include: * a few more sentences about what was changed *...
I agree it would be great to solve this in RocksDB. While it isn't for me to decide, I am not sure sure that will happen given the complexity. For...
To reproduce this I start with this patch: ``` diff --git a/db/db_impl/db_impl_files.cc b/db/db_impl/db_impl_files.cc index da87aa5cd..e011dad50 100644 --- a/db/db_impl/db_impl_files.cc +++ b/db/db_impl/db_impl_files.cc @@ -526,6 +526,10 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {...
For my sake, the call stack to delete an SST after compaction ends is this. However, that depends on [the call to FindObsoleteFiles](https://github.com/facebook/rocksdb/blob/911c0208b917d6e2389f48ad40ec7df58b2c016b/db/db_impl/db_impl_compaction_flush.cc#L2990) finding files that can be deleted, and...
On the bright side, this is explained [in the wiki](https://github.com/facebook/rocksdb/wiki/Delete-Stale-Files). > To prevent it from happening, we take use of a good feature that new files are created using the...
This has compaction IO stats from the end of overwrite for tests that use buffered IO and then O_DIRECT for compaction. The results are from RocksDB version 6.20 First from...
This is an interesting problem. Most of us don't have much experience with mmap for reads. For the usage of MultiGet to read up to ~64M keys in one call,...
Repeating what was discussed elsewhere, benchmark.sh --multiread_batched=true is another way to accomplish this. Seeing this diff reminded me to update my benchmark scripts to do that, so something good has...