Details
-
New Feature
-
Status: Closed (View Workflow)
-
Minor
-
Resolution: Fixed
-
None
-
None
Description
Jira issues created by request per Markus.
The idea is to add a server option which disallows it to ever be elected master. For example, a five node cluster spread across two data centers. Nodes 1, 2, and 3 are in DC1, while nodes 4 and 5 are in DC2. The nodes in DC2 should never be elected master no matter what. An option is needed to add to the server configuration in maxscale.cnf. The possible option discussed was adding a priority of 0 since priority is already used by galeramon to elect the master. This way, a server configured with priority=0 would never be elected master.
Example:
## DC1 - Node 1 |
[dc1node1]
|
type=server
|
address=10.1.1.1 |
port=3306 |
protocol=MySQLBackend
|
priority=1 |
|
## DC1 - Node 2 |
[dc1node2]
|
type=server
|
address=10.1.1.2 |
port=3306 |
protocol=MySQLBackend
|
priority=1 |
|
## DC1 - Node 3 |
[dc1node3]
|
type=server
|
address=10.1.1.3 |
port=3306 |
protocol=MySQLBackend
|
priority=1 |
|
## DC2 - Node 1 |
[dc2node1]
|
type=server
|
address=10.2.1.1 |
port=3306 |
protocol=MySQLBackend
|
priority=0 |
|
## DC1 - Node 2 |
[dc2node2]
|
type=server
|
address=10.2.1.2 |
port=3306 |
protocol=MySQLBackend
|
priority=0 |
In the above example, dc2node1 and dc2node2 would never be elected master, even if dc1node1, dc1node2, and dc1node3 were shutdown gracefully.