Greetings,
First of all, let me say once again what a truly nifty thing samba is.
The performance and functionality improvements with 1.9.17 are just great!
Now for the bad news. ;-) It seems that something has crept into this
new version that causes my Win95 clients to almost always hang when the
user logs out! Here's the specifics:
Server1: PDC, domain logon server, print server
Server2: WINS server
Server3: Read-only file server (PC applications)
Servers 1 & 3 are at 1.9.17p1. Server 2 is at 1.9.17beta5.
Before yesterday all servers were at beta5 with
same problem occurring.
Servers 1 & 2 run FreeBSD. Server 3 runs AIX 3.2.5 .
All servers are in a different IP subnet than clients. All clients are
configured as P-type netbios nodes (non-broadcast). At the same time I
moved from 1.9.16p11 to 1.9.17, the network was changed from a single IP
subnet to this new multi-subnet topology.
Just a few users have home dirs accessible to logon server. With 1.9.16pX
when these few users logged in and out USER.DAT would be correctly
accessed from their home dir.
Now (with 1.9.17) when logging out, the user clicks okay (to confirm the
log-out) the screen dims, the pointer becomes an hour-glass, and progress
halts. The mouse can still be moved, but the system must be reset to
restore operation. This happens during most, but not all, log out
attempts. From looking at a level 3 log, it appears that USER.DAT gets
successfully accessed (if the user has a home dir) during log-in, BUT, the
log shows NO samba activity whatsoever during the log-out attempt!!! (at
least that's what I _think_ I'm seeing).
This behavior is consistent across any of 40 general access lab computers
and any of 40 faculty/staff computers. Given this level of pervasiveness
it would seem that we're looking at some kind of bug.
Note, much of my discussion above focuses on the loading of user profiles.
I'm not necessarily convinced that this is the _root_ cause of this
problem, though problems with this are certainly part of the visible
symptoms. Knowing that the roving profiles code has just been reworked
does tend to make one suspicious. I've also wondered if this could be
somehow WINS related...
What is the best way for me to trouble shoot this? Shall I send
in (to Luke?) level 3 (or higher?) logs? Is anyone else seeing this
behavior? If so, is your topology similar to mine?
Please cc all replies to me directly, as I'm not subscribed to the samba
list.
Thanks in advance for any and all help!
---
-------------------------------------------------------------------------
Charles N. Owens Email: owensc@enc.edu
http://www.enc.edu/~owensc
Network & Systems Administrator
Information Technology Services "Outside of a dog, a book is a man's
Eastern Nazarene College best friend. Inside of a dog it's
too dark to read." - Groucho Marx
-------------------------------------------------------------------------
please cc replies to charles owen: he's not on the samba digest. On Wed, 10 Sep 1997, Charles Owens wrote:> Now for the bad news. ;-) It seems that something has crept into this > new version that causes my Win95 clients to almost always hang when the > user logs out! Here's the specifics: > > Servers 1 & 3 are at 1.9.17p1. Server 2 is at 1.9.17beta5. > Before yesterday all servers were at beta5 with > same problem occurring. > Servers 1 & 2 run FreeBSD. Server 3 runs AIX 3.2.5 .i too am using FreeBSD here at the cafe.> Now (with 1.9.17) when logging out, the user clicks okay (to confirm the > log-out) the screen dims, the pointer becomes an hour-glass, and progress > halts. The mouse can still be moved, but the system must be reset to > restore operation. This happens during most, but not all, log out > attempts.yep, i get this. on _one_ machine, not the other. it's all horribly inconsistent, inconvenient and PISSING ME OFF. for example, i've just removed profiles from my account, deleted local and remote one, and let the w95 machine recreate one for me. what happens? i get to log in, log out, log in, log out x 10 with no hanging of either of the two machines. i even get to run a basic script file (NET TIME \\CB1-GW /SET /YES ^M) more later. i don't like this profile stuff one very much. luke
charles, the only thing that i can suggest is that you try and do what i just attempted, just now. if it doesn't work, i'll let you in on a small secret hack that might have something to do with it (to do with code pages and country codes, in NetUserGetInfo). from a fresh boot-up, instead of logging on to the w95 machine, press escape. then, close progs, log off, and when you get the dialog prompt again, _again_ press escape (this procedure is to _stop_ the w95 machine from thinking that the previous users' profile is the default one. then, backup and then delete the entire contents of the c:\windows\profiles directory. **** NOTE IF SOME OF YOUR USERS HAVE STORED ANY FILES ON THE DESKTOP, YOU WILL BE DOING THEM A REALLY BAD FAVOUR BY NOT BACKING UP THIS DIRECTORY *** then, delete all c:\windows\*.pwl files. then, in each user profile directory on the samba server, delete the user.dat files. i haven't deleted the entire contents of the remote profile directory, and i don't think it's necessary. check the keyboard settings, country settings etc etc. then log in some users, log them out, and see if it creates USER.DAT files. if that doesn't work, then... well. i started throwing mice around. then went upstairs for some hot milk and some salt and vinegar crisps. luke On Wed, 10 Sep 1997, Charles Owens wrote:> Greetings, > > First of all, let me say once again what a truly nifty thing samba is. > The performance and functionality improvements with 1.9.17 are just great! > > Now for the bad news. ;-) It seems that something has crept into this > new version that causes my Win95 clients to almost always hang when the > user logs out! Here's the specifics:
On Wed, 10 Sep 1997, Charles Owens wrote:> Now for the bad news. ;-) It seems that something has crept into this > new version that causes my Win95 clients to almost always hang when the > user logs out!I had this same problem too. All of the machines (except a few) would hang when logging out. If I then sign back on to a particular machine that DID log out, all of the hung machines instantly become unhung. I solved this problem by switching from Samba WINS to NT WINS. Samba WINS appears to have problems with Win95. Later, Jon
I am seeing the SAME problem... but there is no Samba in the picture. I have this problem with NT4.0sp3, Netware 4.11, Netware 4.10, Netware 3.12. I have been working on this issue for several months. I noticed a HUGE increase in this problem this summer/fall when we got in most of our new Win95 B systems. I think there is something in Win95 B which is causing this issue. In 7 of our labs, we did not have this problem when using Win95a, but now with Win95 B it is a huge problem! It also seems to be a bigger problem when there are multiple computers logged in with the same name... I would LOVE to get this resolved (I have about 800 PCs and most of them are having this problem). If anyone else has any idea, PLEASE, lets work together... --------------------------------------------------------------------- | Tim Winders, CNE | Email: twinders@spc.cc.tx.us | | Network Administrator | Phone: 806-894-9611 x 2369 | | South Plains College | Fax: 806-897-4711 | ---------------------------------------------------------------------
Firstly - please don't reply to the reply-to address in this email. I'm working from home at the moment (I'm ill) and haven't got my reply-to's sorted yet. Now, with that out of the way :-). Charles Owens wrote :> Luke wrote: > >> someone's just reported that not using samba for WINS service sorts this >> out. this i find particularly bizarre, but... what the heck! > > !!! Hmmm... along my move from a flat to a subnetted environment, I began > using samba as a WINS server for all clients. So, maybe this _is_ the > area where the problem lies. !!! > > Going a bit further... wouldn't the fact that the locking up still happens > even with my above registry tweak in place seem to indicate that the root > cause of the problem is _not_ really in the area of user profile support? > > Unfortunately I don't have a WinNT server handy to temporarily use for to > test this out (ie. problem goes away when not using samba for WINS > server). I could set one up though, if need be.If you do this, and it cures the problem, then I have a threory as to where the problem might be, and so how to fix it in the Samba source code. Between 1.9.16p11 and 1.9.17 there were changes to the name release code in the WINS server part of nmbd (actually there was a crash bug and a memory leak, but that's by the by :-). If nmbd as a WINS server isn't responding correctly to a name release request then that may cause the hanging. Before you put in a NT WINS server, could you try setting nmbd to run with debug level 10 (that'll produce a *lot* of output so remember to set the log size to zero to allow infinite size logs) and then let a few PC's log out & hang - the resulting trace may be *very* interesting. I *really* want to nail this - if it's an MS bug then that's ok, but I just need to know..... Jeremy Allison. Samba Team.
--On Saturday, September 13, 1997, 12:20 AM +1000 "Luke Kenneth Casson Leighton" <lkcl@switchboard.net> wrote: } } } On Fri, 12 Sep 1997, Bas Laarhoven wrote: } } > At some point I pressed escape to enter windows without domain } > logon, clicked around to the keyboard configuration screen. Then I } > played with the settings a little, made no real changes and } > logged off again. At that point I was able to enter a password ! } } i know. it's a pisser. and this only works sometimes. } } } } ok, summary time: can we bring this one to a close? } Some possibly useful information... There's a "tuning parameter" for TP on Win95 known unfortunately as "MaxConnections". I say unfortunate because that's not exactly what it is. It seems to allocate memory for the IP connection table. Here are some notes I put together after learning about this one from Microsoft Support: Start the Registry Editor (regedit.exe) Select HKEY_LOCAL_MACHINE/System/CurrentControlSet/Services/VxD/MSTCP Add a new STRING value (Edit/New/String Value) Change the default text provided to MaxConnections Double-click on the new entry, and give it a value of 250 Exit from the Registry Editor, and restart the machine. The default value allows a maximum of about 15 connections. After this change, we ran out of memory before the machine wouldn't make any more connections. The total number of connections was 39 when that happened. I also looked at a Win95 shutdown with Netmon. There were four "name release" requests that failed, according to Netmon's decoder. These were: "machine"<03> "machine"<00> "domain"<00> "domain<1e> "machine" is the hostname and netbios name of the Win95 machine shutting down, and "domain" is the name of the NT domain that's used for session authentication by the Win95 box. All failed, according to Netmon, with a "Name Active" error. According to nmbd, these worked (debug level was 3, version is 1.9.17): Name release on name "machine"(20) Name release on name "machine"(3) Name release on name "machine"(0) Name release on name "domain"(0) Name release on name "domain"(1e) It would be good if these were timestamped, but I'm fairly sure they're the ones related to the Win95 machine shutdown. I had to forcibly shut an unnamed process down to get the machine to finish shutdown (not just logout), but after that it *did* shut down. I was surprised - this one mostly hangs during shutdown. It would have been more useful if the shutdown had failed... Jim -- Jim Watt jimw@PE-Nelson.COM Perkin-Elmer Corporation Voice (desk): +1 408 577 2228 PE-Nelson Division Fax: +1 408 894 9307 3833 North First Street Voice (main): +1 408 577 2200 San Jose CA 95134-1701
Any news on this front? How goes the battle? I've been sent a suggestion that I configure my clients to use DNS for Windows Network name resolution, and just circumnavigate the whole WINs thing. Somehow my reading of various sources had left me with the impression that there were some problems with doing this... but I'm going to play with it a bit this afternoon. Any comments as to why this may or may not be a good idea? In any case, of course, I would still very much like to see this nasty nmd bug squashed (at least the current theory seems to be that the fault lies with nmbd). thanks, chuck On Fri, 12 Sep 1997, Luke Kenneth Casson Leighton wrote:> > On Fri, 12 Sep 1997, Charles Owens wrote: > > > Luke, > > > > Just so you know, I've generated a level 10 nmbd log and sent it to Jeremy > > Allison, per his request. I haven't heard from him since, but it didn't > > bounce back to me so I'm assuming he's got it. > > he's ill at the mo :-( > > > If you would also like to see this log, holler. Its big: 1.8 megs! > > agh! > > >--- ------------------------------------------------------------------------- Charles N. Owens Email: owensc@enc.edu http://www.enc.edu/~owensc Network & Systems Administrator Information Technology Services "Outside of a dog, a book is a man's Eastern Nazarene College best friend. Inside of a dog it's too dark to read." - Groucho Marx -------------------------------------------------------------------------