Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I have a Linux fileserver serving up /home for linux and windows users. I was able to connect from my windows client, but not from a DC. Then suddenly I could connect from the DC too.

The linux servers run Centrify clients, and as such are part of the domain. All on same subnet.

This is what the the log.smbd says, repeatedly:

[2010/02/11 11:25:57, 0] lib/util_sock.c:read_data(534) read_data: read failure for 4 bytes to client 192.168.200.3. Error = Connection reset by peer

On Windows it appeared as an "unknown error". EDIT: the error code is "0x80004005".

We are developing a system depended on the samba share, and are worried this will appear again. It would be nice to pin point the root of this.

Any ideas what this might be? Places to look?

share|improve this question
    
What version of Samba and on what distro? I had quite a few problems with Samba and Windows 2008 until I upgraded Samba to at least 3.2. –  Dave Jul 3 '10 at 16:18
    
I wouldn't rely on any solution that blends Windows and Linux. Microsoft are constantly trying to make Windows not work with Samba, to prevent organisations using "free" file servers. –  Garreth McDaid Apr 9 at 13:59
add comment

2 Answers 2

I would blame 'peer', he is the one who reset your connection. There seem to be a lot of theories as to why this problem happens, and how to resolve it.

A couple of places recommend changing the port to 139 instead of port 445

smb.conf
[global]
smb ports = 139

Not sure what version you run, but other sites suggested that recent versions of samba resolve this problem.

share|improve this answer
    
Thank you! We'll try changing the port, and I'll report how it went. It will take a few days until we know if it is working and stable. –  Gomibushi Feb 15 '10 at 7:35
    
Doesn't seem to do the trick, but it seems less frequent. It's a strange mess... We're using Centrify Samba, as we're using this file server for both NFS and samba sharing of home directories. We now have a ticket with Centrify. The only thing we can do to fix this is to clear all caches on the Centrify and samba clients. Might be related to Kerberos. –  Gomibushi Mar 5 '10 at 13:47
    
Case escalated at Centrify. Seems they depend on some packages that were left in Ubuntu 6.06 even for the current 9.10 release! Horrifying. :( –  Gomibushi Mar 8 '10 at 22:28
add comment

[Can't seem to add comment, so am creating a new answer]

Erk. Changing the port to TCP 139 is a fudge. All this is doing is forcing the client to use legacy NetBIOS over TCP/IP (NBT), as opposed to pucker SMB running on TCP 445. Like Dave suggests, check the version of Samba running, and update if poss. Also, check you Windows version on the DC, as Win2k8 R2+ (and possibly R1) require SMB signing by default. This can be turned off, but again, it's another fudge.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.