[clue-tech] Linux-HA and openLDAP problem

[mike havlicek]

--- adam bultman <adamb at glaven.org> wrote:

> I'm setting up a linux-ha 'cluster" with two
> servers.  Both run openLDAP 
> 2.1.
> 
> I'm using the ldap init script to set variables and
> to specify which 
> slapd.conf file I want - if the "balanced" ip
> address exists on the 
> server, then it user slapd.master.conf. If the ip
> address isn't on the 
> server, it uses slapd.slave.conf.
> 
> The failover of linux-ha using heartbeat works fine;
> they restart ldap 
> just as neatly as you please.
> 
> However, my sensing of who is master and who is
> slave isn't working 
> right.  In the ldap init script is set_options,
> which calls ifconfig. 
> pipes things through grep, sed, and awk, and then
> finally returns the 
> result: the IP addreess if it is master, and nothing
> if it is slave.
> 
> Then, there's a bash if statement which tells if it
> has the address or 
> the no value, and then runs slapd with the conf
> file.
> 
> Problem:  I stop a node by stopping heartbeat, and
> the other node kicks 
> in. LDAP is properly restarted with the master
> slapd.conf, and things 
> are fine.  the slave node restarts ldap (I put a
> call to ldap restart in 
> the heartbeat script's stop function)..  I restart
> heartbeat on the 
> server.  The new "master" gets the ip address,  and
> starts slapd with 
> the master conf file.  However, the slave restarts
> ldap, runs ifconfig, 
> and then... notice s that it has the virtual ip, and
> restarts slapd as 
> the master.
> 
> No amountof resetting variables, sleeping, or
> anything like that manages 
> to make the slave restart ldap with the correct
> configuration, and it's 
> really starting to tick me off something awful.
> 
> If you have ideas, apart from "just do master/slave
> and forget it" or 
> "use more sleep calls", feel free to let me know...
> 
> Adam
> _______________________________________________
> CLUE-tech mailing list
> CLUE-tech at cluedenver.org
> http://cluedenver.org/mailman/listinfo/clue-tech
> 
That is the whole idea of clustering. You can do
pseudo clustering using software/hardware tools like
rsync and RAID as well as physically reduntantly
connecting machines using eg crossover cables and
ethernet cards. 

I would be interested in learning about the database
distribution of open ldap in a linux cluster
environment which I beleive was the original question.

There is also the idea of complete failover which I
tend to think of as either mirroring or clonining. On
the other hand with distributed computing you can
"cluster" "services" across your  computing network.

-Mike


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 
_______________________________________________
CLUE-tech mailing list
CLUE-tech at cluedenver.org
http://cluedenver.org/mailman/listinfo/clue-tech