Guillaume Rousse wrote:

Hello list.

I had a look at section 18 of the admin manual, as I was trying to figure difference between scenarios 18.2.2 (N-way multi-master) and 18.2.3 (mirror mode). Whereas the descriptions of both scenarios are quite clear, given examples configuration are quite confusing (at least for me).

First, they are quite hard to compare, as the first one uses LDIF syntax, whereas the second uses the classical file configuration syntax.

Second, they both use the same directives:

• a different serverId for each nodes
• mirrormode on for each nodes
• syncrepl directives for each nodes

The only difference being than for multi-master, each node has syncrepl directives for each nodes (including himself), whereas for mirrror mode, each node only has syncrepl directives for other nodes. But it's hard to figure if that's because of the replication scenario, or just because in the first case, the configuration has to be identical everyhwere given that it's replicated automatically.

So, what is supposed to be different at configuration level between those two scenarios ?

http://www.openldap.org/lists/openldap-software/200910/msg00089.html

Additonal question: doesn't a classical single-master + multiple slaves scenarios, where each slave use chain overlay to automatically forward all changes to the master, also achieve exactly what is described in the mirror mode description ?

Without the mirrormode feature you won't get automatic failover and recovery capabilities.