16 thoughts on “Run multiple garbd arbitrator instances for galera

    • Bernd
    • Arthur Dent the 2nd โ™“๏ธ ๐ŸŒ
    • Stefan Becker
      • It does. It’s 2 clusters but 1 off site monitoring server providing now 2 arbitrators for quorum to handle splitbrain situations of each cluster better. And they are even running different galera versions ๐Ÿคฏ It just works. Really not missing pacemaker/corosync|heartbeat + drbd.


      • That’s an artifact of the spawn process. Systemd starts the daemon and this one forks. In the end it’s running as nobody (or user – handy for ssl certs)

        garbd runs TCP communications but it’s SysVinit script can not deal with multiple instances of the daemon on the same VM ootb.

      • Yeah. Bonus for keeping an eye on cpu/mem usage, auto restart on failure, secure logging and keeping it running without more scripts to check on the process. That’s all ootb with systemd.

        The only monitoring left to do is the cluster/machine health. I roll with Zabbix on this ๐Ÿ˜‡