Once again I'm on the hunt for new knowledge, I'm a part-time DBA as part of my current job, and am particularly interested in any documentation or reading around doing bulletproof multi-master replication topologies - does anyone have any pointers?
In past roles I've implemented and supported such topologies but there are always 'catches' (auto_increment_increment, auto_increment_offset being two big ones) that you only spot at the last minute and which have serious potential to ruin your day.
With a largely InnoDB workload, what are the big problems with multi-master replication and how do you, as a skilled DBA, go about solving them? How does that picture change if you're storing things with MyISAM, or indeed, other storage engines now they're nice and pluggable, perhaps someone has experience with Infobright or another data warehouse?
Emphasis should be placed on recovery techniques for any proposed solution. How does a DBA effectively backup that topology, and how easy is the restore process? Is it bulletproof enough that you can stick a TCP-aware load balancer (hashing on source IP or similar) in front and have zero downtime (or damn close to...) in the event a MySQL master goes pop?
I have read and would highly recommend High Performance MySQL by Baron Schwartz, however, what I'm really after is a couple of really quality websites with all the points covered and links to more in-depth reading material as required. Who's got one of those handy? :)
Bonus brownie points to any solution which can have 'pools' of slaves hung off it for the odd application which has a particularly thrashy read workload.
Thank you very much.