[OpenAFS] Re: about replica

[Cees de Groot]

Martin Schulz <schulz@iwrmm.math.uni-karlsruhe.de> said:
>> I wanted to use AFS storage my all data, including mySQL data, and Oracle data. And the data is always changing.
>
>I think it is a bad idea to load database files in AFS. 
>
[...good arguments snipped...]

Plus there's a better way to run databases:
- Allocate a couple of same-sized partitions on two machines;
- (optional: add network cards and cross cable so there's a dedicated
  synchronization connection);
- Synchronize them with DRBD (http://www.complang.tuwien.ac.at/reisner/drbd/)
- Create filesystem, mount on both boxes;
- Run mySQL on one box, heartbeat (http://www.linux-ha.org/) on both. 
- Go to the beach :-) (ok, Karlsruhe.. Go for a hike in the Schwarzwald)

You can do the same with Oracle; I dunnow how this works performance wise, but
stacking physical partition -> DRBD -> LVM -> Oracle (Oracle can use
partitions for data storage, not?) would be a very flexible solution here.

Of course, the DRBD thingy is neat to have replication of OpenAFS data as
well. I'm toying with the idea of having a separate fileserver partition for
writable volumes, replicated with DRBD and monitored with Heartbeat so if a
box fails, you have automatic fail-over. Dunnow whether this would work with
OpenAFS, though.

-- 
Cees de Groot               http://www.cdegroot.com     <cg@cdegroot.com>
GnuPG 1024D/E0989E8B 0016 F679 F38D 5946 4ECD  1986 F303 937F E098 9E8B
Cogito ergo evigilo