NW3.12 will no longer let W2K Pro machines log in

[andyshriver]

I ran purge and bindfix from our one Win98SE workstation on our 3.12 file server this morning, and, after so doing, all other machines in our LAN (Win2K Pro and Win XP SP2) display the same message: You cannot be authenticated to the SERVER_NAME server because the network path cannot be found. I can still log into and out of the Netware server from the Win98 box and access the resources at will. Please help, as I have 30 people who want to kill me!

Many thanks.

 

[JockMullin]

Hi Andy

Did you shut down/restart the server after doing the bindfix and purge?

If not you may have left a bunch of connections hung.

Check the autoexec.ncf file for the bind commands and make sure using monitor that the bindings have all worked - you may be missing a frame type. Also check the network ID.

If memory serves, bindfix exports the bindery, but doesn't really change it. Bindrest restores the bindery from the exported files. Purge should only have removed deleted files from the server. If there were FAT problems on the SYS volume that may have caused problems if files were cross linked. Have you tried VREPAIR to see if there are any errors (you don't have to write any changes back to the disk)?

What client are you using on the Win 2k workstations? Try forcing the frame type, speed and duplex settings to agree with the server rather than leaving them on AUTO. If the server for some reason does not respond in a timely fashion that often causes NT class OS's not to connect, expecially with the MS Client.

Jock

 

[andyshriver]

Hi Jock,

Thanks for your response. Oddly enough, about 3 hours or so after the W2K and XP users first reported being unable to authenticate, I was able to log in from my W2K machine. Thereafter, everyone else got in with no problems, though some data-related apps were initially sluggish. I am still curious as to why this happened.

The sequence of events was as follows:
1. I logged into the Win98 box, changed to the root directory of the SYS volume and ran purge /a.
2. After completing purge, I ran bindfix from the System directory (no - I didn't shutdown/restart after purge). I believe that, unwittingly, I may have tried to log in before restarting the server, though I cannot be 100% sure.
3. I did eventually shutdown/restart the server, but I was still only able to log in with the 98 box. Foolishly, I ran bindfix again, thereby overwriting the the good reference point from which to run bindrest. I did verify in monitor that the frame types are correct (802.3).
4. Fast-forward 3.5 hours, and I logged in from a W2K box - successfully! We've been up ever since.

How can this be explained? I would like to run purge, bindfix and vrepair in the proper sequence with the shutsown/restarts at the proper times, but I do not wish to relive what I lived yesterday! I would be most grateful to see your take on all of this.

Thanks again.

Andy

 

[JockMullin]

Hi, Andy

That is a bit mysterious, since when you restarted the server it would have closed all connections. The only thing I can think of is the W2K machines still thought they had an active connection. Did you by any chance restart any of them while trying to get them to connect? Just logging off and back on will normally use the same connection. You have likely seen on the server connections for logged off machines are still there with user id NOT LOGGED IN, and after the watchdog timeout limit they will finally disappear.

The 3.5 hrs is consistent with the server forcing an inactive connection to terminate due to inactivity, but resetting the server should have eliminated that. So the short answer is I dunno. What client are you using on the workstations?

If you are running mirrored drives and the purge deleted a whack of files the sluggishness you refer to may have been the disks catching up on the mirroring. if your W2K machines are set to Auto detect the comms parameters then the server may not have been quick enough to satisfy the handshaking resulting in connection failure.

As I mentioned before I always try to pre-define the parameters in NT class workstations because otherwise connecting may be flaky.

Regarding purge: I personally don't use it because then deleted files can't be salvaged and salvage has saved my bacon more times than I can count when a user does something dumb like delete everything in their home directory. The only reasons I can think of are 1. security and 2. to reduce the cache requirements if the server is short on memory or has a lot of small files using up resources. In that case I would set the server to auto-purge deleted files on the fly rather than running a purge all which really ties up the server. Ideally the SYS volume shouldn't have that much volatility anyway unless you have data files there or are doing a lot of spooling.

Vrepair I usually set to run automatically at startup if necessary. It won't repair a mounted volume, so if you need to run it on the SYS volume you will have to shut down to do it. I keep copies of vrepair in the boot partition and on external media (CD or floppy) so it is available if SYS won't mount.

Bindfix I run whenever there have been changes to the bindery and then copy the backup files off into dated folders and/or backup media in case I ever need them for use in Bindrest. I also put copies of the current net$*.dat files there. All my sites have very stable binderies so I don't do it on a regular basis - just once a year whether I need to or not, something like rebooting the server.

Jock