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

Re: multi-master solution?



oberwetter, josh wrote:

Thanks.

Could you also tell me why this functionality is only experimental? Are
there known bugs or limitations?

Other than that someone should review the code and re-label it 'production' ?
AFAIK no. We've been running it for months with no problems.


If modify operations are executed concurrently on the same entry, are the
modifications serialized & thus atomic? If so, which takes precedence?

It's not atomic across all servers (there's no global locking, it's just being done using replication).
It's just like two concurrent write accesses. They are serialized, but you cannot predict which
one succeeds last (overwriting the other writer's changes).

Markus



Josh

-----Original Message-----
From: Markus Storm [mailto:markus.storm@mediaways.net]
Sent: Wednesday, August 15, 2001 3:07 AM
To: oberwetter, josh
Cc: 'openldap-software@OpenLDAP.org'
Subject: Re: multi-master solution?


#define SLAPD_MULTIMASTER 1 in portable.h after configure and before compiling.
(Kurt, mind re-enabling configure --enable-multimaster ?)

Makes slapd NOT write out a replication record on received ADDs/MODs if
the modifying user was the replication uid, thus preventing infinite loops.

Markus

oberwetter, josh wrote:

Sorry for the double post.

To be clear, I saw in the admin guide that there is 

experimental support for

multi-master replication, but I couldn't find enough info to 

play around

with it. More info, please.

-----Original Message-----
From: oberwetter, josh [mailto:joberwetter@grownetwork.com]
Sent: Tuesday, August 14, 2001 6:29 PM
To: 'openldap-software@OpenLDAP.org'
Subject: multi-master solution?


Hi, recently people have discussed load balancing several slave servers in
order to improve response time & redundancy. Can one also load balance the
masters, or will this cause an infinite loop?