Hmm...what you describe below (new replica only picks up new changes,
not
existing content) is specifically not supposed to happen. The code
doesn't allow it
so I'm rather baffled as to how you achieved it. Upon initiating each
replication
transfer, the source server inspects the replica update vector supplied
by the
destination server. In the case of a newly installed server, that ruv
will have a
special null value. Receipt of that value causes the source server to
refuse to send
updates. It prints an error message in the log and basically sulks
forever with respect
to that replica. At least that's how the code is written and that's how
it's always behaved
in my experience.
You might tell us more about exactly how you are 'dropping and
rebuilding'
the replica -- possibly you are doing something strange that results in
an
inconsistency between the server's content and its ruv.
Replica initialization is covered in the docs, here :
http://www.redhat.com/docs/manuals/dir-server/ag/7.1/replicat.html#1104247
Geoff wrote:
Correct guys, I did ask how to add a new master and I understand that
you would destroy the existing replication agreements to insert the new
master.
My test involved setting up 4 way (circular) replication agreements
between 4 DS.
I destroyed one ant the other 3 DS continued to replicate without issue.
I then rebuilt and added the 4 DS back in to the 4 way (circular)
replication.
It did not rebuild itself to contain all of the information the other 3
DS carried and this is what concerned me.
When I added a record to any of the masters it would replicate around
to all 4DS.
But the dropped and rebuilt master never recovered all the data, just
added the new additions.
I notice you mention "initialize the replica". It sounds like a step I
am missing. How does one do that?
Thanks
Geoff
David Boreham wrote:
mallapadi
niranjan wrote:
would it be possible to just install DS and
setup the same hostaname and same replica ID,
and start the replication. Why should we delete the replication
aggreement in other master servers, who had replication agreement with
them.
You don't have to. The way I read your original post, it wasn't clear
that your goal was to
deploy a _replacement_ server (it read like you wanted to add a
completely new server).
Suppose there were more than 4 LDAP server
say 6 and all are Master replicating each other, if suddenly any of the
master server goes down due to hardware reasons, so just installing os
, Directory server and putting the same hostname and replica ID and
starting replication ,
You'd need to manually initialize the replica once (replication from
the peers will
stall until you do this because they'll detect the RUV reset and refuse
to send incremental updates).
would the above be possible ? again
creating
replication aggreements in all the system just because of 1 server
going down.
Yes, this should work just fine.
--
Fedora-directory-users mailing list
Fedora-directory-users@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/fedora-directory-users
--
Fedora-directory-users mailing list
Fedora-directory-users@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/fedora-directory-users
|
--
Fedora-directory-users mailing list
Fedora-directory-users@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/fedora-directory-users
