Here is some log output that occurs on the production during the process of
issuing a net use command from the win95 box to the samba server. The
connection reset by peer message appears immediately after the error 54
network is busy message appears on the windows 95 machine. (This is debug
level = 10)
[1999/05/18 18:21:00, 2] smbd/server.c:main(709)
Changed root to /
[1999/05/18 18:21:00, 3] smbd/oplock.c:open_oplock_ipc(82)
open_oplock_ipc: opening loopback UDP socket.
[1999/05/18 18:21:00, 3] lib/util_sock.c:open_socket_in(721)
bind succeeded on port 0
[1999/05/18 18:21:00, 3] smbd/oplock.c:open_oplock_ipc(110)
open_oplock ipc: pid = 18047, global_oplock_port = 1620
[1999/05/18 18:21:00, 3] smbd/process.c:smbd_process(974)
priming nmbd
[1999/05/18 18:21:00, 3] lib/util_sock.c:send_one_packet(646)
sending a packet of len 1 to (127.0.0.1) on port 137 of type DGRAM
[1999/05/18 18:21:00, 4] lib/time.c:TimeInit(110)
Serverzone is 14400
[1999/05/18 18:21:00, 0] lib/util_sock.c:read_data(384)
read_data: read failure for 4. Error = Connection reset by peer
[1999/05/18 18:21:00, 10] lib/util_sock.c:receive_smb(507)
receive_smb: length < 0!
[1999/05/18 18:21:00, 3] smbd/process.c:timeout_processing(812)
receive_smb error (Connection reset by peer) exiting
[1999/05/18 18:21:00, 2] smbd/server.c:exit_server(406)
Closing connections
[1999/05/18 18:21:00, 3] smbd/server.c:exit_server(433)
Server exit (normal exit)
[1999/05/18 18:21:31, 2] smbd/server.c:exit_server(406)
Closing connections
I have decided to post this message again in hopes that someone might have
a clue as to what is going on. Since the original post, we have also tried
Samba 2.0.4 on the production machine with the same result. Samba 2.0.3
contunues to run on the identically configured test system with no problems
at all.
<From last post...>
I have done a great deal of reading regarding the network is busy error,
but haven't come up with an answer for one particular problem. Here is the
situation. Two identical DGUX machines are sitting on a network. Both
machines are running Samba 1.9.10 with no problems other than performance.
Upon several recommendations about improving performance, they decided to
upgrade to 2.0.3. Downloaded source and compiled it on their test machine.
Worked flawlessly, and it helped the performance problem they were having,
so they decided to put it on their production system. The copied the
binaries from the test system to the production system, and when trying to
connect from various windows 95 machines, they always get the network is
busy (error 54) message when trying a net use command or clicking on the
machine in network neighborhood. They tried compiling the samba 2.0.3
directly on the production server, still had the same problem. They then
reinstalled the old 1.9.10 version and it started working correctly again.
They can connect to the production machine running 2.0.3 from both NT and
other unix machines using smbclient, just not windows 95. The windows NT
does, however, have a long delay when making the initial connection.
Neither machine is running bind, they simply have host tables. We have
tested gethostbyaddr to make sure its not stalling out, and it returns
instantly with either a host name from the host table, or an error if the
name is not in the host table. We have tried having a host table entry for
the win95 machine, and not having a host table entry, and it makes no
difference. The test system does NOT have host table entries for the win95
machines, and it works correctly. Both the test and production servers are
connected to the same fast ethernet switch, all other IP connectivity is
fine between clients and the production server. We have made sure that
there are no other processes running that should conflict with samba.
Also, they are using guest accounts, with security=share, no passwords at
all so there shouldn't be any goofy password or encryption issues here.
The smb.conf files are identical in the global section between the 2
servers, only the shares themselves are different. The smb.conf file is
very plain, no socket options, no NT domain stuff, no security, nothing.
Only the workgroup name and dns proxy=no are in the global section. This
has become a pretty serious issue, and we obviously need some additional
help, because we are rapidly running out of ideas. Please let me know if
you have any thoughts of what we might be overlooking. Thank you in
advance.
-----------------------------------------------------------------------
W. Bryan Caudle
President
Commonwealth Technical Services, Inc.
Voice: 804-639-5400 Fax: 804-739-7007
http://www.ctsi.net
