Le 29/12/2015 11:04, Quanah Gibson-Mount a écrit :
--On Tuesday, December 29, 2015 10:39 AM +0400 Jephte Clain jephte.clain@univ-reunion.fr wrote:
Le 28/12/2015 19:12, Quanah Gibson-Mount a écrit :
I'm not sure to understand what you are saying. the "master" and the "slaves" are identical, and the cn=config are the same on all hosts so they are truly replicates, right?
No. Replicas do not accept writes. Replicas do not have a master configuration for cn=config. Replica's do not have server IDs.
ok I guess I understand. this is the reason why I usually call them "slaves", not replicas (but I messed things up and called them replicas this time ^^) I also have replicas that only replicate data (or a subset of data) for some services
the slaves are there in case of catastrophic failure of both masters (we had one of these failure for another service due to a problem with the shared storage. No one want to have this kind of emergency...) If the master(s) crash, I just have to choose a slave as the new master, slapcat the cn=config database, update the provider address, slapadd the updated config, and update the loadbalancer settings. this is a bit of work but at least we can restore service in a (relatively) small amount of time.
In fact, the configuration is generated by a script I wrote several years ago, and all databases for a "master" configuration are generated with a syncprov overlay and a syncrepl directive, to enable "seed replication":
After the problems I had, I took the time to read the admin guide again and noticed the accesslog database was not to be replicated, or it seemed so I then wondered if the replication I configured on the accesslog database was the cause for my issues... hence my question to be sure I understood correctly
Accesslog is unique to a given master.
Ok that's what I wanted to know for sure
Shouldn't the doc stat this clearly?
Here is another question: if an accesslog database is stricly local to a server, how should two masters in mirror mode be configured? I have a bi-master setup with an active/passive configuration: the loadbalancer only send the requests to the first master, unless it stop responding. if the first master crashes, and writes are diriged toward the second master, won't I lose the accesslog informations?
Every master must have a unique server ID. Each master will replicate the writes from another master, and update their accesslog accordingly. You will not lose any writes.
that's a relief
I'm guessing your configurations are generally incorrect.
Yes, I'm updating them right now to disable replication of the accesslog
Thanks a lot for the clarification. In case you come to the reunion island someday, I owe you a beer!
best regards, Jephté CLAIN