Lightly changes are made to provider an optional "force-switch" when manually failover the Mysql cluster.
Reason
The decision would be taken for some scenarios as below:
I have auto-failover running as well as I want to manually switch the master by some time. And I'd like it to be a "forced-switch" rather than following the original configuration.
Even if the delay is high, "no_master" is made by MHA manager itself(as the new_master I pointed to, might be a "oldest server"), I would have the opportunity to ignore it, as long as the "recover_new_master" works.
Changes
An option input for ignoring the "no_master" keyword, so that my target new master is always guaranteed to be chosen for a new epoch
Additional global variable "new_master_ignore_no_master" will have effect on the decider whether "candidate is bad or not"
A field and "Set" method is added to "ServerManager.pm"
Abstract
Lightly changes are made to provider an optional "force-switch" when manually failover the Mysql cluster.
Reason
The decision would be taken for some scenarios as below:
Changes