[Date Prev][Date Next] [Chronological] [Thread] [Top]

RE: Adding attributes (in a massive way)....


Are you saying the complete database data replacement method is really
needed to be sure the slave/master are synchronized? What is your experience
with inconsistencies? Obviously you have no problem with that at present.
How often were you dealing with such problems earlier? My special concern is
the slurp behaviour in case one of the slaves is not on-line for a
significant time. In theory all update should be replicated to it when it is
back, but if there is an intensive load of provisioning updates (say 0,2-0,3
per sec) and the slave does not come up within 15-20 hours. Is the only
limitation the partition size allowed for the replog file? 

-----Original Message-----
From: Andrew Bacchi [mailto:bacchi@rpi.edu]
Sent: 19 March 2003 20:42
To: openldap-software@OpenLDAP.org
Cc: Peter Marschall
Subject: Re: Adding attributes (in a massive way)....

I agree with Frank on this method.  I have changes each day to a 14K
user directory, and I replace the whole database each evening in a
matter of minutes.  I therefore have a perfectly synchronized
master/slave pair.  

One difference between Frank's method and mine, I don't do the slapcat,
I build an ldif file from a global /etc/passwd.

On Wed, 2003-03-19 at 14:03, Frank Swasey wrote:
> Today at 3:44pm, Peter Marschall wrote:
> > I cannot believe tthe numbers for the ldapmodify.
> > Did you actually measure them ?
> I did not measure a full 54000 entry modification.  I have sat for HOURS 
> and watched an ldapmodify spew forth one "updating xxx" per second when 
> I made massive changes (defined as run a slapcat, process with a perl 
> script to generate the ldif to update everyone, run ldapmodify on that 
> ldif).  I have long since given up doing things that way.  
> Now, instead of writing a perl script using Net::LDAP to bind, 
> search/update (all entries), unbind.  I write a perl script using 
> Net::LDAP to read the ldif file generated by slapcat, make the updates, 
> write a new ldif file, then I run slapadd to replace the whole 
> directory.  It has proven to be just plain faster in my environment.
> It's been about six months since I even tried the "live update" process 
> for anything more than a few entries.
> -- 
> 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 ===
Andrew Bacchi
Staff Systems Programmer
Rensselaer Polytechnic Institute
phone: 518 276-6415  fax: 518 276-2809