[Date Prev][Date Next]
Re: NFS with OpenLDAP [different topic, similar question]
Today at 8:38am, Chuck Pierce wrote:
> Similar topic, different question; just no nfs.
> Say I have shared disk between two servers. If one server fails, the other
> server picks up the disks (and starts slapd). Would that cause data
> corruption problems? Would it be better to run slarpd and have the
> secondary (standby) server change from the slave to the master?
Depending on what the master was doing at the time it died, the files
could definately be corrupted before the second server ever touches
them. It's certainly more reliable, doesn't have a single point of
failure (if you lose the shared disk...), to use slurpd between two
independant servers and promote the slave to master (that can be done
automatically). It's the bringing the busted machine back into the
picture that can cause problems and must be designed for.
> Also, if I do use a slave server, and the slave server goes down. How to
> I ensure that the slave it "up to date" when I bring it back online? Do
> I need to copy the database from the master to the slave, or does slarpd
> take care of it for me?
slurpd will save all the updates and run them when the machine comes
back up. The issue then becomes, how recent is the backup of the
directory you had to restore on the slave after you fixed the hardware
problem and whether the updates that have been lost (between the time
the backup was taken and when the failure occured) cause the waiting
updates to fail.... if data integrity is critical to you, then I would
think that you would want to stop slurpd and migrate the database from
the master to the slave before bringing the slave back up. Of course
with a 2.0 server that (usually) means taking the master server down to
run slapcat (or having a second slave you can copy the database from).
Frank Swasey | http://www.uvm.edu/~fcs
Systems Programmer | Always remember: You are UNIQUE,
University of Vermont | just like everyone else.
=== God Bless Us All ===