nmbdlost.1.slim7721@spamgourmet.com
2005-Jun-15 23:37 UTC
[Samba] nmbd fails to recognize network change locked on old DMB
After setting up a working SAMBA test server, attempts to move it to production failed. The plan was to replace an old file server upgrading to 3.7 OpenBSD and SAMBA 3.0.14a. The test server worked in a test environment, but when moved to the production network replacing the existing server, SAMBA failed. The failure centers around SAMBA's inability to recognize that it has changed networks with different IPs. It reports the old DHCP server as Master Browser and fails to elect itself browser even though the server it is subordinate to no longer exists. I have done away with the cache, read multiple pages of documentation, tweaked the smb.conf file all to no avail. i found one older listserver article that mentioned files /var/samba/cache/browser.log and cache.log but I could not find those files. When I query SAMBA with the tools it reports old network stuff (addresses and browser) as if it has a history file and it is not going to change regardless of the new environment. ie SAMBA fails. Details: Old environment: About 2.2 OpenBSD with a really old SAMBA. i386 Pentium PC Server 4 workstations 1 3com Office Connect Firewal/Gateway(does DHCP) to cable modem and Internet. 192.168.200. network addressing Comments: Solid as a rock never a problem. Boss wants a new PC with XP. Admittedly old workstation giving problems. Test Environment: 3.7 OpenBSD and 3.0.14a SAMBA i386 Server Duron processor 1 new Dell PC workstation running XP i386 Diskless (runs off CD) Dell PC as an internet gateway,DHCP server 1 Windows 98 PC (that I never attempted to have join the domain) 1 Windows 95 workstation 10.10.11. network addressing Obtained CD's for OpenBSD 3.7 and installed i386 binary to test server. FTP downloaded SAMBA 3.0.14a and made the install from scratch. Installed the XP workstation and got it working with SAMBA. The Windows 95 workstation was also able to join the domain. All is fine. Moved the test server and XP workstation to production environment. Changed ip addressing from 10.10.11 to 192.168.200 and booted up the server and XP workstation. This is where things went whacky. Both machines were set for DHCP client. No serious errors were reported, but nothing worked. The SAMBA logs for the workstation reported "connection denied from 192.168.200.47" the nmbd.log said "There is already a domain master browser at IP 10.10.11.130 for workgroup chaz.com registered on subnet UNICAST_SUBNET. Packet send failed to 10.10.11.130.....". I have rebooted. Eliminateed cache and reduced time to live in my SAMBA server configuration. What I need is a startup parameter for nmbd that says "FLUSH and RESET UNICAST_SUBNET"? Looking at the man pages for smbd, smb.conf, nmbd I do not find any such parameter. I have done away with DHCP on the XP and server boxes and static addressed them with table in hosts file. We can ping each other by name or address fine. SAMBA will not get off the UNICAST_SUBNET error. In researching the problem on one of the SAMBA listservers someone wrote that if you want to know what SAMBA was doing look at /var/samba/cache/browser... and another file with the same path(not sure i have the names correct, I tossed the slip of paper when they were not found). I looked around and could not find those files. Not much of anything on the net about UNICAST_SUBNET. I assume that this might be an internal SAMBA dataname or field. I do not have any desire to go to the source code. It seems to me that in an election for master browser server, if there aren't any other candidates, the one surviving candidate should win. Is this a SAMBA bug? Am I just ignorant? Had the Linux gateway bid and received Master Browser Status in the test environment (apparently so)? Did OpenBSD bow in humble, memorial reverence to Linux and prevent SAMBA from doing its job? Will someone give me a parameter to tweak or a file I can whack? I would rather not rebuild SAMBA again.