Red Hat Bugzilla – Bug 253100
Samba Master Browsing not working correctly
Last modified: 2007-11-30 17:07:46 EST
From Bugzilla Helper: User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8.1.6) Gecko/20070725 Firefox/2.0.0.6 Description of problem: I have several RHEL 4 ES and RHEL 5 servers. We recently upgraded 6 of our file servers to RHEL 5 (samba-3.0.23c-2.el5.2.0.2) using the exact same 'smb.conf' files that were on the server when it was RHEL 4 (samba-3.0.10-1.4E.12.2). This problem is isolated to the RHEL 5 servers. The following entry is logged repeatedly and since one of the servers that got upgraded was the domain master browser (DMB) all Windows browsing is basically broken. [root@ampere ~]# tail /var/log/samba/nmbd.log [2007/08/16 13:44:10, 0] nmbd/nmbd_browsesync.c:find_domain_master_name_query_fail(351) find_domain_master_name_query_fail: Unable to find the Domain Master Browser name ESG<1b> for the workgroup ESG. Unable to sync browse lists in this workgroup. The problem seems to be that each of the RHEL 5 servers wants to be its own master browser instead of letting the designated DMB be the master browser. [root@ampere ~]# smbclient -NL ampere Domain=[ESG] OS=[Unix] Server=[Samba 3.0.23c-2.el5.2.0.2] Sharename Type Comment --------- ---- ------- ampere Disk Ampere File Server hr Disk Human Resources Directory qbdata Disk QuickBooks Data Directory IPC$ IPC IPC Service (ampere) Domain=[ESG] OS=[Unix] Server=[Samba 3.0.23c-2.el5.2.0.2] Server Comment --------- ------- AMPERE ampere Workgroup Master --------- ------- ESG AMPERE Version-Release number of selected component (if applicable): samba-3.0.23c-2.el5.2.0.2 How reproducible: Always Steps to Reproduce: 1. Install 'samba-common-3.0.23c-2.el5.2.0.2', 'samba-client-3.0.23c-2.el5.2.0.2' 'samba-3.0.23c-2.el5.2.0.2' on two RHEL 5 servers 2. Designate one of them as a DMB in /etc/samba/smb.conf [global] workgroup = esg server string = esg security = SHARE password server = None guest account = esg name resolve order = wins host socket options = IPTOS_LOWDELAY TCP_NODELAY IPTOS_THROUGHPUT load printers = Yes printcap name = /dev/null show add printer wizard = No preferred master = Yes domain master = Yes wins support = Yes remote announce = 10.10.12.210 10.10.40.200 10.10.60.200 10.10.80.200 10.10.110.200 remote browse sync = 10.10.12.210 10.10.40.200 10.10.60.200 10.10.80.200 10.10.110.200 socket address = 10.10.10.200 create mask = 0644 guest ok = Yes hosts allow = 127.0.0., 10.10. 3. Designate the other server to be a WINS client in /etc/samba/smb.conf [global] workgroup = esg server string = ampere security = SHARE password server = None guest account = esg name resolve order = wins host socket options = IPTOS_LOWDELAY TCP_NODELAY IPTOS_THROUGHPUT load printers = No printcap name = /dev/null show add printer wizard = No wins server = 10.10.10.200 remote announce = 10.10.10.200 remote browse sync = 10.10.10.200 socket address = 10.10.10.205 create mask = 0644 guest ok = Yes hosts allow = 127.0.0., 10.10. Actual Results: Each RHEL 5 becomes its own DMB and none of the Windows clients see them in their browse list. Expected Results: Windows browse list should match output from 'smbclient -NL <DMB>'. Additional info:
The error indicate that the servers can't find the domain master browser not that they are one. I need some more info to understand if this is a configuration problem or a bug. Just to be sure can you do the following: 1) - Check that you can actually retrieve the DMB from the WINS server nmblookup -R -U<WINSserver> 'ESG#1B' 2) - add local master = yes to all servers that need to be local master - add domain master = no and preferred master = yes to all servers except the domain master - raise the log level of nmbd of one of the servers that fail and attach the nmbd.log
1) Output from both the WINS/DMB server and client were the same: name_query failed to find name ESG#1b 2) 'nmbd.log' file is attached.
Created attachment 161705 [details] NMBD Log File
(In reply to comment #2) > 1) Output from both the WINS/DMB server and client were the same: > > name_query failed to find name ESG#1b If this is the case then there is a problem with the wins server. Can you provide the WINS server nmbd.log file ?
Created attachment 161714 [details] 'NMBD' log file for WINS server
I've attached the 'nmbd.log' file for the WINS server.
This log shows that the main server becomes a domain master browser. Can you run the above nmblookup command on the WINS server directing it to itself and verify if it resolves?
I ran it 3 different ways and it all produces the same resutls. Please note the name of the DMB/WINS server is 'esg.esgrp.net' and the workgroup name is 'ESG'. [root@esg ~]# nmblookup -R -U esg 'ESG#1B' added interface ip=10.10.10.200 bcast=10.10.10.255 nmask=255.255.255.0 Socket opened. querying ESG on 10.10.10.200 Negative name query response, rcode 0x03: The name requested does not exist. name_query failed to find name ESG#1b [root@esg ~]# nmblookup -R -U esg.esgrp.net 'ESG#1B' added interface ip=10.10.10.200 bcast=10.10.10.255 nmask=255.255.255.0 Socket opened. querying ESG on 10.10.10.200 Negative name query response, rcode 0x03: The name requested does not exist. name_query failed to find name ESG#1b [root@esg ~]# nmblookup -R -U localhost 'ESG#1B' added interface ip=10.10.10.200 bcast=10.10.10.255 nmask=255.255.255.0 Socket opened. querying ESG on 127.0.0.1 name_query failed to find name ESG#1b Samba is allowed through the firewall, but I tried the command with an without 'iptables' running and it made no difference. I also disabled selinux's controls of Samba by setting the following booleans: smbd_disable_trans --> on nmbd_disable_trans --> on
The thing that is odd is it actually is the master brower: [root@esg ~]# smbclient -NL esg Domain=[ESG] OS=[Unix] Server=[Samba 3.0.23c-2.el5.2.0.2] Sharename Type Comment --------- ---- ------- IPC$ IPC IPC Service (esg) Domain=[ESG] OS=[Unix] Server=[Samba 3.0.23c-2.el5.2.0.2] Server Comment --------- ------- ACCT-01 AMPERE ampere ARCHIVE archive BACKUP backup ESG esg GAUSS gauss ICEHAWK-CSK JNU jnu MAXWELL maxwell RDM rdm SCADA scada SCANNER SOS sos VCR vcr VOLT volt Workgroup Master --------- ------- ESG ESG
If you run that same command without the #1B is does respond with a name lookup. [root@esg ~]# nmblookup -R -U esg 'ESG' added interface ip=10.10.10.200 bcast=10.10.10.255 nmask=255.255.255.0 Socket opened. querying ESG on 10.10.10.200 Got a positive name query response from 10.10.10.200 ( 255.255.255.255 ) 255.255.255.255 ESG<00>
Yes there seem to be a problem where the WINS server actually wins the DMB elections but somehow forgets to set that into the WINS database itself, I need to dig in the code to find out what's going on. Meanwhile we have new packages in the beta channel you may want to try out, they may fix your problem. If you do, can you please report abck if that fixes your problem?
I'm not familiar with the beta channel, can you give me a link? The searching I've done on Google groups would lead me to believe this may have been is v3.0.24 but I didn't see anything technically significant as to what the issue actually was.
(In reply to comment #12) > I'm not familiar with the beta channel, can you give me a link? > > The searching I've done on Google groups would lead me to believe this may have > been fixed in v3.0.24 but I didn't see anything technically significant as to > what the issue actually was. I meant "been fixed in v3.0.24"
(In reply to comment #12) > I'm not familiar with the beta channel, can you give me a link? it's an rhn channel, which you can subscribe to using the 'alter channel subscriptions' link from the system's page on rhn > The searching I've done on Google groups would lead me to believe this may have > been is v3.0.24 but I didn't see anything technically significant as to what the > issue actually was. In the beta we have 3.0.25 that's why I am asking if you can try that out.
I tried the beta version (samba-3.0.25b-0.el5.4) and the results are the same. The servers still do the exact same thing.
Please can you provide me the nmbd.log file of the DMB/WINS server at log level =10 ? Please do a restart and let it log at that level for 5 minutes Also set max log size = 0 during this tjme so that logs don't wrap. (log level 10 can slow doiwn the server considerably, if it is a production system better to do these test off peek hours).
Created attachment 161899 [details] 'nmbd.log' file at log level = 10
I've attached about 5 minutes worth of the 'nmbd.log' file with the "log level = 10" parameter set for the WINS server.
Thanks, I am trying to reproduce it.
Tim, I have reproduced the problem. It is a configuration issue. "socket address" is the culprit. If you use it nmbd will not listen on 127.0.0.1 and it will not be able to send to itself the WINS registration packets. The socket address option is meant to be used only in very specialized environment where you have multiple smbd running on a multi-homed system. In those cases usually nmbd is given a different configuration. Can you remove socket address and report if this fixes your problem?
Removing the "socket address" from the configurations on the RHEL 5 servers does appear to have corrected the issue. Is this a bug or is just a misconfiguration? This was not a problem in RHEL 4 (samba-3.0.10-1.4E.12.2). In any event, thank you very much for your help.
I'll tag it as a misconfiguration, normally to restrict the interfaces you listen to you can use: "interfaces" and "bind interfaces only" and the documentation warns you to always include "lo" or localhost or bad things happen. "socket address" is a very obscure and stricter thing that allows you to restrict smbd to a single specific address. But this is generally unwanted and should be used with care and in environments where a multihomed system is used to run smbd with "different" configurations on different ip address. Thanks for bringing this up, it will help others if they spot the same problem.