From: Thomas S H. <tha...@gm...> - 2011-03-10 04:44:38
|
I have simplified the ucarp, but set up a daemon that manages the metalogger because after a failover the metadata on the loggers was corrupt until there was a fresh pull from the master. I don't have a document yet, as you can see by my github, I have been very busy lately, but so far on my tests this setup has been flawless, even with rapid cascading failovers, downtime between failovers depend on the speed the drives storing the metadata, I have some ideas to speed this up, but life happens. I know that failover is in very high demand for MooseFS, if anyone wants to help with this patches and github pull requests are VERY welcome. The code is here: https://github.com/thatch45/mfs-failover Let me know what you all think! P.S. I have a few other projects that might be helpful for people, if you are interested take a look at salt, clay and varch, we use them extensively at Beyond Oblivion |