IP static/dynamic & domain login (W98)

[jdharm]

I won't get into the details of the catastrophe yesterday, but the part that pertains to this problem is that all the DHCP servers in the domain went down. For some reason every server (5 NT, a couple of which were running Mac services, and one FreeBSD) on campus (various locations) rebooted "spontaneously" at the same instant. The resulting cluster***k kept me here till two in the morning.

The problem I am having is with the Win98 machines. The NT and 2K workstations and the Macs all took exception to the absence of the DHCP servers, but they recovered when the servers came back online. A large portion of the Win98 workstations didn't fare as well. For some reason about 1/3 of the machines would boot to a Windows logon rather than a domain logon. When I went into Windows I could surf the net, but I couldn't browse the local network or see any other computers in the workgroup or domain or get the policies and config files from the server, but I could ping the servers from a command line. I tried a myriad things, but the only thing that made them work properly was to assign static IP's. Release/renew cycling when DHCP didn't fix the issue and going back to DHCP after manually assigning the IP would revert back to the problem.

A curious facet of the problem. If I went into the network properties and OK'ed the settings (while dhcp) without changing anything the computer would boot up to a domain login screen. Reboot and the domain/local network was gone again. It would stay gone till I had gone to the network properties and OK'ed everything again, after which it would be good for one boot. While everything was working as it should you could log any number of users off and back on, but at the next reboot you were without the domain/local network again.

Any ideas at all what this is about? I seem to recall something similar here before but I couldn't find it. I have it patched together now with 55 machines on static IP's, but ideally we would like them back on DHCP.

Josh

[i]Sorry folks, I'm an idiot. Moose out front shoulda told ya. [i]

[feelsick]

to change from booting to network to a windows logon requires one of two things to have happened;
1. they have been re-configured to do so - unlikely - but you can check in network control applet for primary logon...or more likely
2. they can't see the network. I know that you can force them to communicate by giving them a static ip but its not quite the same thing. on power on they look around and try to join their workgroup. in a domain environment the workgroup as specified in network properties is often an ignored detail, which is ok mostly but whatever is set there is actually the first domain a win98 tries to become a guest of.
the other thing about 98 is if set for dhcp but can't get an ip it uses the last settings it had. this might include subnet mask so check that is right or clear.
firstly start,run,winipcfg and release all.
then physically power them off and on - this last point is v.important.

[jdharm]

Re: 1.

Nothing changes when I OK the settings and everything works and nothing changes when rebooting to the problem. All settings are the same as they were last week when everything was fine.

Re: 2.

All settings are correct. I monitored the DHCP manager while the machine I had just forced to release it's address rebooted. I saw it successfully take out a new lease in a completely new scope and still couldn't "see" the network despite being able to access the Internet.

Re: power cycle

To be honest I don't recall having pulled the power cord on anything. I did hit the power button (AT machines), but I never pulled a cord other than the LAN patch cables. I'll give it a whirl. I don't expect anything, but I've seen stranger things.

The "now it works, now it doesn't, now it does, now it doesn't" thing is what's throwing me.



Josh

[i]Sorry folks, I'm an idiot. Moose out front shoulda told ya. [i]

[feelsick]

I think you have provided your own answer. If it gets an ip from a different scope it must believe it is part of the subnet which that scope is for. It does seem like the memory effect of 98 clients and a physical power off might work. Also as I said earlier, making the workgroup name the same as domain is good practice. clear out all caches , any mapped drives (esp reconnect ones), and why not use lmhosts file?

[jdharm]

The problem started while using the old scope. We changed scope to start the whole domain fresh (the original scopes on the two DHCP servers overlapped a bit).

Making the domain name and workgroup name the same and assigning both is/has been standard practice.

Not sure exactly how to use the hosts file in this way, but I'll learn and give it a shot.

We're really hoping that the problem will sort itself out this week when the domain clients get settled into the new scope and the WINS and DNS servers get updated/synchronized. We'll see. I wanted to try to track down the source of the problem in case it persisted.

Josh

[i]Sorry folks, I'm an idiot. Moose out front shoulda told ya. [i]