Multiple updates have been made across CockroachDB to improve resilience and performance under adverse conditions. Together, the features below along with other smaller enhancements, we’ve achieved a 20% performance improvement in this release.
Network partitions between a leaseholder and its followers heal in less than 20 seconds, since the leaseholder no longer needs to heartbeat a single node liveness range.
Outages caused by liveness failures last less than 1 second, since liveness is now determined by a store-level detection mechanism, not a single node liveness range.
Performance is equivalent (within less than 1%) to epoch-based leases on a 100 node cluster of 32 vCPU machines with 8 stores each.
Spin up your first cluster today or speak with an expert.