Tek-Tips is the largest IT community on the Internet today!

Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!

  • Congratulations strongm on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

Master Replica Mess

Status
Not open for further replies.

Datong

IS-IT--Management
Feb 15, 2005
118
GB
This is a strange situation I have not seen before. The scenario is : In one area there are 4 Nov5 SP6 servers. Each server has 2 Masters on it and several RWs. The odd thing is that the masters are identical. What appears to have been done is the server was originaly setup with one parition on, then a second was added with a different name but the objects within the original are aliased into the new parition. This has resulted in 2 masters of the same partition on the same server - which is the same for each of the 3 other servers. I dont understand the thinking behind this and need to be sure if I start removing duplicate masters that things will not start to fall apart. The reason is to reduce the amount of replication going on as the servers are not particulalry hi-spec and the network is sluggish as a result. Any advice would be appreciated. PS the old partition is not completely redundant.
 
when you say partition are you meaning the o is alias'ed
like someone had dsmerged them
 
Okay I don't quite understand what you mean. but. in general, when you have a mess on your hands, you need to cover a few bases first.

- make sure time is in synch on all servers. do a dsrepair -->time synchronization, make sure they all say yes and there are no errors. If there are, get those resolved first, then see if any of the wierd stuff you see clears up on its own.

Regarding the multiple masters.. you can have multiple Replicas on one server, that's not a big deal. But you should not have multiple masters replicas of the same partition. that is bad. So if that's what you're saying is the case, you need to correct this. I would opt for not removing the replicas completely at first, just changing them to READ/WRITE replicas and making sure each replica ring only has one master.

Marvin Huffaker, MCNE
 
Terry - I dont think DSMerge was a factor, from what I have established they decided on a Name changing exercise which was obviously done badly. The O's are not aliased but the Server object and volumes, plus some users are. I was confused at first because I thourght there were 2 servers involved but there is only one in each case.
Marv - I have checked timesync and done a DSrepair and everything appears ok, although a little slow as you can imagine. One question, if I have 2 master replicas on one server, how do I know which is being accessed via dsrepair or is it generic?
Thanks guys, at least I can start to resolve this with your help.
 
Hit another snag. I cant demote one of the masters without promoting an RW on another server to a Master. This sort of defeats the object by placing another master elsewhere. I think my options are severly limited with what I can do here. Is there a way to remove a master without promoting another because if I do promote one I would be in the same predicament as before when trying to remove the new master. What would happen if I merged the two partitions, would the two replicas merge and the one I want rid of be demoted or deleted or would it just synchronize into one Master ? Ive not used dsmerge before.
Thanks.
 
Hi,

The first thing I would say here is to slow down and get a better picture of the entire situation. If you start messing with master replicas without knowing exactly what is going on then you could get yourself is some real trouble.

I am not sure that I fully understand the entire situation. In your first post you mentioned 2 separate partitions (one which had aliased objects), so if that is actually the case then the Master replicas can both be valid and they are NOT exactly the same.

This would be tough to diagnose without a much better idea of the exact layout of your tree, all the paritions, and all the replicas.



gwexpertslogo.jpg

Gioffre Consulting
Check out our Novell Training Cruises
 
It's very possible that what you are seeing is completely normal and you just don't understand the way partitions and replicas work. I would recommend calling in a 3rd party and getting their take on the situation. You could turn a very normal situation into a complete disaster in a few bad moves.

If you really want to understand partitioning and replication, build a test tree with multiple containers. Use 2-3 servers so you can see the replica behavior. Create some partitions at variuos containers, and add replicas. You'll get a better idea of how it should work, and when you kill it, you won't get fired and sued.

Marvin Huffaker, MCNE
 
Hi, sorry I have not replied sooner, but Ive been away. Anyway, back to the thread. Marv - I understand enough to know that the setup here is not right. The practice of multiple masters and replicas seems to be inherent throughout the entire tree here. Its a complete nightmare. The network is so slow and now I know the reason why. Unfortunately (or fortunately for me) It is not something I can correct anyway as they are planning to downgrade to windows 2003. Thanks all for the advice and support but I think I can close the lid on this one. Oh well, off to build a Win2k3 server. Oh joy !
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top