yugabyte-db
https://github.com/yugabyte/yugabyte-db
C++
YugaByteDB is a high-performance, transactional database for building distributed cloud services. It currently supports Redis API (as a true DB), Cassandra API and SQL is coming very soon.
Triage Issues!
When you volunteer to triage issues, you'll receive an email each day with a link to an open issue that needs help in this project. You'll also receive instructions on how to triage issues.
Triage Docs!
Receive a documented method or class from your favorite GitHub repos in your inbox every day. If you're really pro, receive undocumented methods or classes and supercharge your commit history.
C++ not yet supported75 Subscribers
Add a CodeTriage badge to yugabyte-db
Help out
- Issues
- [Docs] Export YSQL logs from YugabyteDB Aeon clusters
- [YSQL] [Table Locks] FK check re-acquires the referenced table lock for every row; after 256 lock requests in one statement each acquire is a tserver round trip and a WARNING log line (100k-row FK insert 11.75 s vs 1.19 s without FK)
- [YSQL] Parallel Append over a partitioned table: each worker rebuilds every partition's relcache entry from the master on every execution (~6 catalog reads per partition per worker); default plan 7x slower than serial at 600 partitions
- [YSQL] Autocommit single-row SELECT ... FOR UPDATE runs a distributed transaction: 5x tserver CPU and 6x Raft RPCs of a fast-path single-row UPDATE
- [YSQL] Single-row primary key SELECT deep-copies the read request once per tablet (ClonePgsqlOps), CPU grows with tablet count
- [YSQL] Every idle backend wakes ~20 times per second (pggate reactor coarse timers): 0.066% of a core per idle connection
- [YSQL][Table locks] Every idle connection keeps a live transaction that heartbeats its status tablet twice a second (~0.26% of a core per idle connection)
- [YSQL] Connection starts duplicate and queue behind catalog prefetches already running on the same node
- [YSQL] GROUPING SETS returns scalar-subquery value instead of NULL for omitted grouping keys
- [YSQL] Index scans recheck in PG the index conditions they cannot bind, even when DocDB could filter on them
- Docs
- C++ not yet supported