[KLUG Members] Server Mirroring

[Adam Tauno Williams]

> I'm looking for a solution opensource or comercial to mirror data on two
> servers in real time.
> Something like Vinca standby server.
> My final goal is to have a fault tolerant solution with a database.
> Have the database mirrored to two server.
> Use HeartBeat to fail over the actual database server application if one
> fails.
> I plan on having several services running on BOTH systems.
> If all the data is mirrored to both systems then the services can be run on
> either system.
> Anyone know if any such mirroring software?

If by "database" you mean an RDBMS (SQL database engine): No.  ***Reliably***
and ***consistently*** replicating an RDBMS is really really really really
really H-A-R-D.  If you absolutely need this as a hot-pair.... buy DB2.

Otherwise you can hook two systems to a RAID cabinet and have one take over if
the other dies; but you don't get both at the same time and they need to be
within feet of each other (unless you can afford fiber, which is more than
DB2).

MySQL and PostgreSQL both have hideous hacks to add brute force replication with
very shaky atomicity,  It'll work for an application like a web forum, or maybe
a job tracking system, but I wouldn't put it within a thousand feet of a
business system (unless you think loosing a bunch of writes to the Accounts
Receivable module is OK,  those guys in suits might disagree - strongly).

Honestly,  unless you have ALOT of data and thousands of users, you are better
of with something like an x440 with internal redundancy.   I have IBM servers
with RAID, dual power supplies, temperature monitoring, ECC RAM, etc.... they
just don't go down.  Put it in a cabinet with its own ventilation (like the
vent on your dryer) two UPSs and two cooling units, and your about as safe as
your going to get. [Assuming your building has a sprinker system,  the cabinets
are pretty sealled when closed].   Use 1U nodes like the x300's for processing,
 you can add redundancy cheap and easy at that level;  but replicating the
database itself is just really freakin' hard.  And a bunch of redundant servers
all sitting in the same room.... are all going to bake or drown at the same
time.

If you're talking about files, then perhaps GFS?