The FoundationDB team released version 6.0.15 of its distributed, NoSQL database, yesterday. FoundationDB 6.0.15 explores new features such as multi-region support, seamless failover management, along with performance changes, and bug fixes.
FoundationDB is an open source, multi-model datastore by Apple that lets you store multiple data types in a single database. All data in FoundationDB is safely stored, distributed, and replicated in the Key-Value Store component. FoundationDB offers high performance on commodity hardware, helping you support very heavy loads at a low cost.
Let’s have a look at what’s new in FoundationDB 6.0.15.
- FoundationDB 6.0.15 offers native multi-region support that dramatically increases your database’s global availability. This also offers greater control over how failover scenarios are managed.
- Seamless failover is now possible in FoundationDB 6.9.15, allowing your cluster to survive the loss of an entire region without any service interruption. These features can be further deployed so that clients experience low-latency, single-region writes.
- Support has been added for asynchronous replication to a remote DC with processes in a single cluster. This improves the asynchronous replication provided by fdbdr as servers can fetch data from the remote DC in case all the other replicas have been lost in one DC.
- Additional support has been added for synchronous replication of the transaction log to a remote DC. This makes sure that the remote DC need not contain any storage servers.
- The TLS plugin has been statically linked into the client and server binaries. There is no longer a need for a separate library.
- The fileconfigure command has been added to fdbcli which configures a database from a JSON document.
- The master recovery time for clusters with large amounts of data has been significantly reduced.
- Recovery time has been significantly reduced for cases where rollbacks are executed on the memory storage engine.
- Clients can now update their key location cache much more efficiently after the reboots of storage servers.
- Multiple resolver configurations have been tuned to carry out the job balancing work more efficiently between each resolver.
- Clusters that been configured to use TLS would get stuck, leading to all their CPUs getting used for opening new connections. This issue has been fixed now.
- The issue of TLS certificate reloading causing the TLS connections to drop until the processes were restarted has been fixed.
- The issue of Watches registered on a lagging storage server taking a long time to trigger has been fixed.
- The capitalization of trace event names and attributes has been normalized.
- Memory requirements of the transaction log have been increased by 400MB.
- The replication factor in status JSON has been stored under redundancy_mode instead of redundancy.factor.
- The metric data_version_lag is replaced by data_lag.versions and data_lag.seconds.
- Several additional metrics have been added for the number of watches and mutation count and are exposed through status.
For more information on FoundationDB 6.0.15, check out the official release notes.