Details
-
Type:
Task
-
Status: Open
-
Priority:
Major
-
Resolution: Unresolved
-
Fix Version/s: None
-
Component/s: None
-
Labels:
Description
GTID allows to automatically fail-over a master by promoting a slave as the
new master, bringing it up-to-date with any other slaves and switching
existing slaves to use the new master. The procedure is described in the
documentation.
https://mariadb.com/kb/en/global-transaction-id/#start-slave-until-master_gtid_posxxx
We need a script that implements this procedure. The user can then run this
script to promote a slave as a new master, simply by supplying connection
information for all involved servers.
Gliffy Diagrams
Attachments
Activity
- All
- Comments
- Work Log
- History
- Activity
- Transitions
Possible to add an option to set last executed GTID in the information_schema.servers replicate with corosync cluster API any node could decide who is the most up to date slave in a consistante state, proxy and connection pools could enter the group member to make proper switch over without losing any command. Connection and SQL not in the group member could be refused.
Check this out, golang so no dependencies like perl/python/ruby needs at install:
https://github.com/tanji/mariadb-tools
Blog post:
http://varokism.blogspot.co.uk/2015/06/slave-election-is-welcoming-gtid.html
Could be possible to use the mysql.server table ? Could auto populate an information schema table when a slave connect and update a last seen date to produce warnings and errors in case of a outdated slave is picked by the command.
I would rather see a command more than a script as. One would like to create a corosync cluster plugin to be loaded inside MariaDB, synchronize all server information_schema.servers table with the corosync Totem Single Ring Ordering and Membership Protocol and can call such failover command on the Synchronization Engine plugin