[Samba] Connection reset by peer / The network is busy
chad at rhiannonweb.com
Fri Mar 26 20:50:20 GMT 2004
I have seen posts concerning this in the archives, but with no real
solution. I've seen it first-hand twice now.
I am running Debian Woody, with us3.samba.org's Samba 3 sources added to my
apt-sources. 3.0.2a was the first version of Samba installed. The
SMB.conf is used off of a previous "Test" system, then updated to reflect
changes in the network as it stands now.
The server is fairly bare, with just minimal packages installed. It's only
use is for file-sharing and one Access database sharing. (Soon it will
also host a MySQL database, hence the horsepower.) It is on a Tyan 2880S
motherboard with dual Broadcom GbE ethernet cards, TG3 module. Kernel is
2.4.25-K7-SMP custom, Samba is 3.0.2a
I went to copy files over from our old server to the new one via the network...
System 1 (pdr, Win98SE) Select files, copy, select new directory on new
server, paste. Files copy, re-map network drive, everything's happy.
System 2 (My computer, Win2k Pro) Same as above. Slower because my laptop
is a piece of garbage, but it's always been that way, even copying local to
System 3 (Server, Win2k Pro) Same as above for "programs" share, used for
network copies of various programs we use. (ACT, MS Office, etc.) 5.7GB
worth of information. Was doubling as a test to make sure the two systems
can see each other, as the old server will be used as a jury-rigged
nightly-updated mirror in case of panic.
System 4 (lab, Win98SE, Via Rhine II NIC) Same as above. Gets about 1/2
way through copying files (800MB of Excel files), and the following occurs:
Client beeps: "Cannot create or replace [filename]: The network is busy.
Server log: "[2004/03/26 15:06:23, 0] lib/util_sock.c:read_socket_data(342)
read_socket_data: recv failure for 4. Error = Connection reset by
At which point, the client needs to be completely rebooted before it will
re-connect to the server.
Added the new server to the WINS list of the client, and started using a
domain logon. Error persisted.
On this last try, I decided to monitor the progress with smbstatus while
the files copied over. First time I ran smbstatus, the client beeped, and
the error occured immediately, much sooner in the process than normal.
I've tried everything I can think of. We don't have problems of this
nature with that system connecting to the old server (Win2k Pro. We were a
tiny company then.) I would just break it up into smaller sections, but I
would rather fix the problem than workaround if possible. Right now I'm
moving over less-used information. However, we are stretching the limits
of the old server rapidly, and need to finish the transition soon.
The last time I noticed this problem was using Samba 2.999 on Xandros
1.0. Same network, using it as a testbed. Reformatting the system and
re-installing Xandros from scratch worked like a charm. Purge / re-install
of Samba did not. However, at this point a purge/reinstall of the OS is
not an option. Going back to Xandros is also not an option, as it was
causing other problems.
Also, last time stress-testing Samba by streaming an MP3/OGG playlist would
cause this several times a day. I tried the same earlier with no problems
from pdr. I'm not ruling out a network problem, but it would be nice to
find out more about the specific error so I know where to focus my
server string=%h server (Samba %v)
print command=lp -d %p %s; rm %s
# invalid users=root
max log size=
passwd program=/usr/bin/passwd %u
passwd chat=*EntersnewsUNIXspassword:* %nn *RetypesnewsUNIXspassword:* %nn .
obey pam restrictions=yes
short preserve case=yes
name resolve order=lmhosts wins bcast host
add user script=/usr/sbin/useradd -d /dev/null -g 100 -s /bin/false -M %u
veto oplock files=/*.mdb/*.xls/*.mp3/*.ogg/
passdb backend = tdbsam guest
username map = /etc/samba/users.map
interfaces = eth0
write list=[userlist removed]
valid users=[userlist removed]
More information about the samba