This morning at 5:05am, our master load-balancer stopped being the master and the backup took over and then gave back control to the master. Two problems: 1) the master shouldn’t have stopped and 2) the master should have taken control back when the backup gave it back. Fortunately, a quick restart of the keepalived daemon fixes it.
I checked the mailing list archives for the load-balancing software and, of course, the problem was discovered this week and there’s lots of talk about it. Hopefully, a fix will be made to it soon. Fortunately, it only happens every week or so.