Chris Phillips wrote: > > > On Tue, Jun 9, 2009 at 4:06 PM, Rich Megginson <rmeggins at redhat.com > <mailto:rmeggins at redhat.com>> wrote: > > Chris Phillips wrote: > > Hi, > > I've a cluster of boxes with replication form two multimasters > to 6 read only replicas. There appears to be a problem in the > replication in that the error logs state that the DSA is > unwilling to process updates for a specific user account, so > the replication status in the idm just stays at saying it > started rather than completed. I could just delete the account > and recreate it, but as it's unfortunately *my* account (and > is in this state *possibly* because I was messing with the > resetpasswordretrytime field (or something very similarly > named) which I get the impression is treated differently to > other fields) I'd like to avoid deleting the account. > > To this end I'm hoping a suitable solution is to remove > whatever the change is that is trying to be pushed across, but > I can't see any way with SSL replication to see what the > actual attributes it doesn't like are. Any way to pull this > straight out with ldapsearch or something? Any tips for > elegantly troubleshooting this in a heavily locked down > environment would be appreciated. > > > Yes, it probably has to do with one of those password related > operational attributes. There are a couple of ways to handle this > 1) change your replication agreement to exclude the attributes > passwordRetryCount, retryCountResetTime, and accountUnlockTime - > you do this by adding these attributes to be excluded in > fractional replication - you should be able to modify your > existing replication agreements to exclude these > 2) add the attribute passwordIsGlobalPolicy in cn=config to "on" > on your servers - this will allow those attributes to be replicated > > > This seems to fit in exactly, thanks. If I set this value on a read > only replica, what will happen if it is locked out on that replica? > Presumably despite this setting that can't get replicated back up to > the multimasters? Correct. You can set up chain on update to have "global" lockout - see http://directory.fedoraproject.org/wiki/Howto:ChainOnUpdate > > As an alternative to changing the policy, can I manually undo these > changes? TBH I'm not too clued up on what triggers an attribute like > this to be chosen to be replicated in the first place, if it's a > hidden timestamp or such. Yes. Those attributes are operational attributes - you have to ask for them explicitly in the ldapsearch request. You should be able to set them manually as directory manager. But for now, the problem is that the changes are in the changelog and the server is attempting to replicate them. I suggest setting the isglobal attribute in the replica, allowing the change, then disabling the replication of those attributes and disabling global policy. > > Thanks > > Chris > > ------------------------------------------------------------------------ > > -- > 389 users mailing list > 389-users at redhat.com > https://www.redhat.com/mailman/listinfo/fedora-directory-users > -------------- next part -------------- A non-text attachment was scrubbed... Name: smime.p7s Type: application/x-pkcs7-signature Size: 3258 bytes Desc: S/MIME Cryptographic Signature Url : http://lists.fedoraproject.org/pipermail/389-users/attachments/20090609/aac62115/attachment.bin
