Peter Dillinger
Peter Dillinger
Summary: ... when malloc_usable_size is known at compile time to be available. This feature has been around for a couple of years and users haven't reported any problems with it....
C++20 has at least one feature we would really like to use: std::span for passing (or in some cases, returning) constant references to sequences of objects regardless of whether they...
Currently, we have code that calls ReopenWritableFile on the latest WAL file on recovery: https://github.com/facebook/rocksdb/blob/9.0.fb/db/db_impl/db_impl_open.cc#L1555-L1566 with the purpose of freeing any fallocated space that might be leftover. This is unnecessary...
Iterators are commonly used with prior known bounds, and even if those bounds are specified to the iterator, there are potential optimizations RocksDB can do in a batch "fetch all...
Summary: Upon discovering that LockWAL() does not block file ingestion, I looked into whether its behavior should be expanded to block that or whether it should block as little as...
Summary: Not sure if this change is desirable given the unit test churn, but the unit test is not clear, in part because the meaning of readahead options is unclear...
Summary: Follow-up from #12757. New infrastructure to DEBUG builds to catch certain ReadOptions being ignored in a context where they should be in effect. This currently only applies to checking...
Both `adaptive_readahead` and `readahead_size` describe RocksDB behaviors, but not how that option (or others) affects that behavior. The comment for `adaptive_readahead` refers to a non-existent option `max_auto_readahead_size.` I don't see...