Samba / Windows 2000 problems
Good news - we've just converted the school over from an NT PDC to a gnu linux/samba one and things are almost working perfectly. We have about 30 acorn clients which continue to work perfectly, and similarly about 80 nt4 workstation clients which work fine (but with an unpleasant bodge I'll mention later) and about 30 W2K clients which almost always work perfectly. All the windows clients are working with policies and mandatory roaming profiles, and astonishingly, they seem to work :) Sometimes, however, when we log in at one of the W2K machines it will complain that it cannot connect to the shares on the server, and starts up with a local default profile instead. If we log out and try again it works fine. The problem is not related to server or network load - it can happen when the network and server are idle but for the W2K login. Has anybody come across this? Any solutions, or clues for some tests to do? Now for the unpleasant bodge :) The only way we've been able to con2prt to work for setting up the default printer for the machine, is to add users to the group set up for domain admins (domain admin group in smb.conf). A similar problem arises with connecting to shares - we can't, unless the user is in the domain admin group. These problems occur within the context of the logon script. Anyone got any ideas? Cheers -- Phil Driscoll
On Sat, 2 Mar 2002, Phil Driscoll wrote:
Sometimes, however, when we log in at one of the W2K machines it will complain that it cannot connect to the shares on the server, and starts up with a local default profile instead. If we log out and try again it works fine. The problem is not related to server or network load - it can happen when the network and server are idle but for the W2K login. Has anybody come across this? Any solutions, or clues for some tests to do? Now for the unpleasant bodge :) The only way we've been able to con2prt to work for setting up the default printer for the machine, is to add users to the group set up for domain admins (domain admin group in smb.conf). A similar problem arises with connecting to shares - we can't, unless the user is in the domain admin group. These problems occur within the context of the logon script. Anyone got any ideas?
What happens when you try to connect outside of the logon script? (This shouldn't make a difference, but it might be easier to see any error messages.) What do the Samba logs say? You can find them in somewhere like /var/log/samba, and the log level will be determined by the smb.conf parameter "log level" or "debug level". HTH, Michael
On Saturday 02 March 2002 6:46 pm, Michael Brown wrote:
What do the Samba logs say? You can find them in somewhere like /var/log/samba, and the log level will be determined by the smb.conf parameter "log level" or "debug level".
We've tweaked the logging level upwards and when a machine fails to login, we get this in the logfile: ************* [2002/03/04 08:15:50, 1] smbd/service.c:make_connection(610) langlab_02 (10.13.2.52) connect to service Profiles$ as user guest (uid=503, gid=101) (pid 18176) [2002/03/04 08:16:32, 0] smbd/oplock.c:request_oplock_break(1026) request_oplock_break: no response received to oplock break request to pid 18173 on port 34884 for dev = 900, inode = 104334 for dev = 900, inode = 104334, tv_sec = 3c832d3c, tv_usec = 9ac64 [2002/03/04 08:16:32, 0] smbd/open.c:open_mode_check(555) open_mode_check: exlusive oplock left by process 18173 after break ! For file pupils/SendTo/My Briefcase.lnk, dev = 900, inode = 104334. Deleting it to continue... [2002/03/04 08:16:32, 0] lib/util.c:smb_panic(1055) PANIC: open_mode_check: Existant process 18173 left active oplock. ************* I've disabled oplocks on all shares and the problem seems to have gone away. I still don't understand why this problem occured, and was only present on the W2K clients. I suspect that for performance reasons we ought to have oplocks enabled on the home directory shares, but clearly, reliability is more important than performance. Cheers -- Phil Driscoll
Recently I read in one of the PC mags that by default Win2K and Win-XP both disable low-level Ethernet packet decoding / processing in hardware on the network card. MS have set this to be handled in software by the PC's CPU instead (sorry I can't remember which mag - PC-Pro, Pc-Plus or PCW - I'm moving house so I can't easily lay my hands on this mag right now). There are some advantages to this (I forget what), but on a network segment with PCs attached to a hub as opposed to a switch (as in a typical classroom) this means every Win2K PC has to do a lot of CPU intensive background processing just identifying all the packets on a network segment, almost all of which it's ultimately going to reject anyway! Now maybe this processing overhead (especially at a time when a whole class is logging in at the same time) is slowing down the PC to such an extent its causing the login to time out and the machine to default back to local machine only mode. Incidentally, it is possible to disable this 'feature' in Win2K. Maybe someone else can dig out this reference for you. As far as I can recall it's also referred to somewhere deep inside MS's knowledge base.
I still don't understand why this problem occurred, and was only present on the W2K clients. I suspect that for performance reasons we ought to have oplocks enabled on the home directory shares, but clearly, reliability is more important than performance.
David Bowles Education Support
participants (3)
-
David Bowles
-
Michael Brown
-
Phil Driscoll