users have 4 min. login from 1 partition to another

[ajack]

We have users in one partition on our tree mapping a drive to a NW5.1 in another partition. The authentication takes four minutes. The client and target NW.5.1 server communicate in seconds. The hold up seems to be between the target NW5.1 server and clients Home NW5.1 server. We see a lot of IP traffic between these servers, once the servers finish the target server finishes the authentication. Other than creating a read/write replica is there any other options or ideas that might the problem. Where should the replica reside (target or home) server?

Thank you

 

[SimonMag]

Are the servers part of the same tree? or are the clients trying to authenticate to the target server in another tree?

If the servers are in the same tree, I would suggest having a read/write replica on the target server and the Master on the host server. Also make sure that the replicas are in sync. (Run a DSREPAIR and check syncronisation status of all replicas)

This in theory should speed up the process.
Hope it helps.


Simon Magee
Network Engineer
Bottomline Technologies Europe

http://www.bottomline.co.uk

mailto: smagee@bottomline.co.uk

 

[TheBoySmith]

With you saying in another partition, as smagee suggests partition placement should help. I'd look in nds manager and check what sub-refs each server has, this may help in deciding what partitions need to be added. Lee Smith
Xenon Network Services
Snr. Technical Support
LSmith@xenon-uk.co.uk

 

[DanMc]

Even if you need to make a more efficient tree design, it should not take 5 minutes unless you have a VERY slow network or VERY large tree. You are probably timing out on some operation. You didn't specify what your protocols and clients are, so this may not apply at all. According to Novell, NT/2000 clients perform asynchronous lookups to all available name services and do not use the "name resolution order" setting.

I had a similar problem caused by strange default behavior build in to outdated Win9x clients. We noticed the slow logins were occuring only when our ISP was down. Our DNS was hosted on our ISP's network. Even though we had all our Netware traffic on IPX, and explicitly configured Client32 to use IPX as the default protocol, we discovered that the client was trying to do a DNS name lookup of the server (and other SAPing hosts) before performing an IPX login. It was waiting for a response (which never came) for over 5 minutes. No matter how we configured the clients they would still try to perform the lookup and wait. (including the Client32 props for name resolution order, specifying NDS and SAP before DNS or SLP for lookups)

Upgrading out clients to 3.1SP2 solved the problem. Then when 3.2 came out, the problem reappeared, so we downgraded to 3.1SP2. In 3.3 the problem is fixed again. In addition, we have moved our DNS on-site.

The next troubleshooting steps I would try:
1.Make sure you're using the newest client
2.Assign DNS names to the servers to match their NW names
3.Sniff the wire to see if the client is repeating some kind of request during the delay period.
4.Monkey around with the name resolution order settings on the client if it's Win9x. Use what makes sense for your network.