Uploaded image for project: 'MariaDB MaxScale'
  1. MariaDB MaxScale
  2. MXS-2885

Group slaves with failover



    • New Feature
    • Status: Closed (View Workflow)
    • Major
    • Resolution: Won't Do
    • None
    • N/A
    • binlogrouter
    • None


      If the binlog router will be phased away, please, consider this FR for its replacement.

      MaxScale has the feature of secondary masters, allowing it to fail over between Galera nodes in reads from when replicating one Galera cluster (Cluster A) to another (Cluster B). Unfortunately, only one node from Cluster B can be attached to MaxScale as its slave (since in Galera we have log_slave_updates always on).

      This FR asks for the ability to tag slaves (that connect to MaxScale) into virtual groups so that when multiple slaves from the same group are connected, always only one will receive the actual binlog stream and the remaining slaves will receive "no new events". This will allow MaxScale to transparently start feeding the binlog to another Galera node on Cluster B when the one that received it before goes down.

      Combining his FR with secondary masters should allow one MaxScale instance to read from any Galera member on Cluster A (with transparent failover) and send to any Galera member on Cluster B (with transparent failover), resulting in a resilient and self-maintaining async link between two (remote) Galera clusters (which, when set up in the opposite direction also, will result in two Galera clusters being in active-active replication with 100% self-maintenance).


        Issue Links



              toddstoffel Todd Stoffel (Inactive)
              assen.totin Assen Totin (Inactive)
              0 Vote for this issue
              1 Start watching this issue



                Git Integration

                  Error rendering 'com.xiplink.jira.git.jira_git_plugin:git-issue-webpanel'. Please contact your Jira administrators.