Make the deployment of the database better service multinode
Kolla-kubernetes is using mysql as a database. The community needs to investigate the functionality of mysql in the multinode environment vs galera using shared storage (likely ceph).
Kubernetes is a different management tool than normal. Because of the way Kubernetes abstracts the presence of a service, it guarantees that if there is a node available, the database will be active. This concept is called cloud-native.
Some of the options to consider are: whether to use galera or mysql or whether to have a database per service, ect...
Blueprint information
- Status:
- Not started
- Approver:
- Ryan Hallisey
- Priority:
- Undefined
- Drafter:
- Ryan Hallisey
- Direction:
- Needs approval
- Assignee:
- None
- Definition:
- Discussion
- Series goal:
- None
- Implementation:
- Unknown
- Milestone target:
- None
- Started by
- Completed by
Related branches
Related bugs
Sprints
Whiteboard
I think there should be two different axis here.... mariadb instance(s) backed by ceph is one axis. It can be a cluster of 1 ... N. This makes it more fault tolerant when N > 1. Galaera provides this. The other axis, is sharding the db's. you don't have to share a db instance between nova and neutron for example. we should be able to launch multiple database instances (clustered or not) as well. A user could minimize their admin overhead and get semi HA by just launching a db per service without clustering. Or some sites may want to do both clustering and sharding - kfox1111
Note, the discussion also applies to rabbit - kfox1111