Hello, there are some examples of n-way replication configuration examples around and I would like to ask for some authoritative configuration advice. test050 for example gives the impression that serverID parameter has to be multivalued at each configuration file, i.e. in a 3-way replication:
serverID 1 $URL1 serverID 2 $URL2 serverID 3 $URL3
same applies to replication ID, it seems that sid and rid have some mutual relation, and all participating hosts should be aware of all rids, i.e. in a 3-way replication:
syncrepl rid 1 provider=$URL1 syncrepl rid 2 provider=$URL2 syncrepl rid 3 provider=$URL3
mirrormode true overlay syncprov syncprov-reloadhint true syncprov-checkpoint 1 5
with this configuration each host has to connect to itself and to all participating hosts. I have done some extensive testing in the last few days and found that a system configuration as described above caused contextCSN to be set to sid 000 and led to a quite unstable system. The following configuration provided correct contextCSN setting and a stable system:
serverID 1 syncrepl rid 1 provider=$URL2 syncrepl rid 3 provider=$URL3 mirrormode true overlay syncprov
serverID 2 syncrepl rid 2 provider=$URL1 syncrepl rid 3 provider=$URL3 mirrormode true overlay syncprov
serverID 3 syncrepl rid 3 provider=$URL1 syncrepl rid 1 provider=$URL2 mirrormode true overlay syncprov
My question now is: is this a proper way to setup n-way replication or is this a wrong, but working, configuration.
-Dieter
Dieter Kluenter wrote:
Hello, there are some examples of n-way replication configuration examples around and I would like to ask for some authoritative configuration advice. test050 for example gives the impression that serverID parameter has to be multivalued at each configuration file, i.e. in a 3-way replication:
serverID 1 $URL1 serverID 2 $URL2 serverID 3 $URL3
test050 configures replication on both the cn=config tree and a user database. As such, all server configurations will be made identical once replication starts. So, they all need to know about all 3 servers at the outset. If you're not replicating cn=config, then this isn't necessary.
same applies to replication ID, it seems that sid and rid have some mutual relation, and all participating hosts should be aware of all rids, i.e. in a 3-way replication:
No. As the documentation clearly says, rids only need to be unique within a particular slapd configuration. rids in one server have absolutely no relation to rids in any other server.
openldap-software@openldap.org