You are viewing an older version of this section. View current production version.
5.1 Release Notes
Role Based Access Control
MemSQL 5.1 features the new Role-Based Access Control (RBAC) enhancement for companies with comprehensive security requirements.
MemSQL Enterprise Edition enables organizations to leverage the power of roles. To enable this feature, two new objects were added to the security model: roles and groups. Roles are collections of permissions and groups are collections of users. Roles are then applied to groups and users are put into groups. Customers can create their own roles such as:
- Security Officer - managing users and passwords
- Cluster Administrator - managing the MemSQL cluster
- Backup Operator - performing backups
- Application User - executing per-application DML statements
MemSQL also provides management functions that determine the set of permissions for a given user.
Learn more about the new RBAC functionality in the RBAC Security Deployment Guide.
Maintenance Release Changelog
2016-07-15 Version 5.1.3 (LATEST)
- Fixed an issue with caching buffers released by query execution (sorts, distincts, etc.) even when the server is under memory pressure or when more then 25% of memory is already cached. Fix prevents Out of Memory error.
- Fixed an issue with INSERT … ON DUPLICATE KEY UPDATE … VALUE(..) queries crashing after a DROP INDEX operation is run on the table they’re running against.
- Fixed a rare crash in the columnstore background merger under a heavy write workload.
- Adds user control of sync timeout for replicating master data. Important for maintaining Disaster Recovery capabilities over slow networks.
2016-06-29 Version 5.1.2
- Fixed several minor bugs around RBAC
- Prevent configuration issues by disabling creation of non-unique hash indexes
- Change default maximum memory settings: maximum_memory defaults to 90% of physical memory instead of 100% even when swap is available.
- Modified recovery behavior to trigger GC earlier
- Fixed a bug where certain queries including LIMIT in a sub-query of a distributed join could return wrong results