List:Replication« Previous MessageNext Message »
From:Rick James Date:May 15 2009 8:09pm
Subject:RE: advice on master-master replication failover support
View as plain text  
> > * "Automatic" is not possible.
> 
> Well, it's working with mmm for me.

The killer is a partial network outage.

The monitoring cannot connect to the hot Master, so it decides that the
master is down, and signals the other Master to take over.  But the
first Master is still alive.  Then you end up with two machines, each
thinking it is the only master.  Now data is flowing into each one.
That's a mess I never want to face.

Note further, for MM to be of much use, the masters need to be in
geographically distant places so that no flood, earthquake, etc can
bring down both Masters.  This aggravates the chance of a "partial
network outage", and requires the monitoring to also go long distance.
Thread
advice on master-master replication failover supportCarl Marcinik13 May
  • RE: advice on master-master replication failover supportRick James13 May
    • Re: advice on master-master replication failover supportMarcus Bointon14 May
      • RE: advice on master-master replication failover supportRick James15 May
Re: advice on master-master replication failover supportCM14 May
Re: advice on master-master replication failover supportCM14 May
Re: advice on master-master replication failover supportCM14 May
Re: advice on master-master replication failover supportCM14 May
Re: advice on master-master replication failover supportCM14 May
Re: advice on master-master replication failover supportCM14 May