Machines lose access to a workgroup

[JohnConnolly]

I have a problem with machines losing access to a workgroup consisting of 18 machines: 14 XP Pro, 1 XP Pro-64, 2 Linux Samba servers and one NAS. There are 3 switches (2 at 100 Mb/s and 1 at 100/1000 Mb/s with all machines on the same subnet (192.168.100 mask 255.255.255.0). Netbios over tcp/ip is enabled on all machines.

Everything works fine for a while then a machine loses all access to the workgroup. When I try to open the workgroup or any share on a "locked out" machine, access is denied. If I search for machines using * as a wildcard, no machines are found.

I thought I had solved this when I found a problem on a Samba server but it has just come back again. Have I got too many machines for a workgroup? I have seen many comments on internet saying a workgroup shouldn't be too big but nothing definitive.

I'm going crazy - any help much appreciated!

 

[xwb]

It isn't how many machines you have on the network: it is how many machines are accessing one machine.

Unless you are using a server, the number of connections to single machines is very limited. I'm running a network with 10 machines. Everyone can see the server but they can't always see each other. The only way to transfer files round is to go through the server.

 

[JohnConnolly]

Does the "how many machines are accessing one machine" apply if the machine being accessed is a Linux Samba server?

More info:

The machine that is currently locked out is XP Pro x64. If I reboot I can access the workgroup for a few minutes then everything slows down and finally it's locked out again. If I use "net use" in those first few minutes I can map a share to a local drive which works fine for a while, then slows down, then fails.

 

[smah]

Unless it has a hardware issue, the samba server should easily handle connections from 18 Windows machines. Do I understand correctly that this happening on only 1 particular machine? Check for errors in the event viewer - Start, Run, evenvtvwr.msc

 

[JohnConnolly]

Yes, only one machine - the one running x64 edition. Can't see any problems in the event viewer. Just tried forcing the master browser onto the samba server - doesn't seem to make much difference except I can now open the workgroup but can only access the shares on the two samba servers and on the local machine.

 

[smah]

How is name resolution handled on this network? If you force NetBIOS over TCP/IP for this machine, do things improve? Network connection, Properties, TCP/IP, Properties, Advanced button, Wins tab.

Also double check for minor no-no types of things like duplicate computer names on the network, same username & computername, etc. (But I would expect these to at least show in event viewer if they existed)

 

[JohnConnolly]

NetBIOS over TCP/IP is already forced on all machines. There are no duplicate computer names or same username / computername. There are no network problems - ping works fine, even big ping (ping computer_name /n 10 /l 32000) has no lost packets with an average of less than 1ms (gigabit connection).

 

[JohnConnolly]

More Info: About 15 mins after I forced the master server onto the samba server the x64 machine regained access to the workgroup and the shares but with strange behaviour. I decided to concentrate on just one server, a NAS (really a cut-down samba server with all the software on a flash module). Sometimes clicking a share opens instantly then clicking a folder maybe the same, maybe takes 10 minutes! Same thing going back up through the tree! I can see no pattern - just seems to be randon even for reopening the same folder. All other machines are normal - everything on the NAS opens instantly. I'm beginning to wonder if there's some wierd problem with XP-Prox 64 Edition...