[Samba] Samba Upgrade issues

[Collins, Kevin]

Over the last three days I've been upgrading my Samba infrastructure.  This involved moving from Red Hat Enterprise 3.0 (Samba 3.0.9) to Ubuntu 5.10(Samba 3.0.14) and some new hardware.  For the most part things went well.  But I do have some unresolved issues that I would like to get some feedback on.  Keep in mind that this entire setup has been working properly for more than two years in this fashion.

First a bit of ASCII art:


            ---------------
            | Main Office |
            |    PDC      |
            | Master LDAP |
            ---------------
                  |
                  |
                 VPN
                  |
                  |
        ----------------------
        |                     |
        |                     |
 ----------------      ----------------
 | Office No. 1 |      | Office No. 2 |
 |     BDC      |      |     BDC      |
 |  Slave LDAP  |      |  Slave LDAP  |
 ----------------      ----------------


In the "Main Office", we run a 60/40 spit of machines running Windows XP and Windows 2000, leaning heavier toward XP.  One laptop (running Windows XP) gave us problems logging onto to the domain for about 20 minutes.  After a minor change to the LDAP configuration and a restart of Samba on the PDC, this machine came online.  The remaining machines in this office came online with very little issues - the only issue being a slow logon the very first time.

In "Office No. 1" every machine runs Windows 2000 and everyone of them had to removed and re-added to the domain before logons would work.  We kept getting errors stating that the domain controller was unavailable or the computer account password in the domain was incorrect.  These errors happened immediately on the windows clients and nothing was recorded in the Samba logs.

In "Office No. 2" we are running Windows 2000 on one machine and Windows XP Pro on all other machines.  The Windows 2000 client exhibited the same symptoms as described in Office No. 1.  One of the Windows XP clients exhibited the same symptoms as well.  The remaining XP machines worked fine.  To cure the troublesome XP client, we had to remove the machine from the domain, delete the LDAP computer account and then rejoin the domain.  After that process everything seems to be functional.

The upgrade process went like this:  On Friday of last week, we had every user turn their computer off as they left for the day.  We left all of the servers online through the weekend.  On Monday, we upgraded the PDC and checked a few workstations to make sure that things were OK.  On Tuesday we were involved in getting the rack in the server room buttoned up - no changes with the exception of a machine or two being taken offline for a few minutes while cables were routed.  On Wednesday, we upgraded the Office No. 1 BDC, handled the problem with the Laptop in the Main office and then Upgraded Office No. 2 BDC.  Because of the problems seen in both of the remote offices, this morning, we went to every workstation in the main office making sure that they functioned properly.

So my question is why did we have the problems in the remote offices?  Why could they not contact the domain controller?  Why would a removal and rejoin cause the problem to go away?  Should I be worried about future occurrences of this phenomena?

--
Kevin L. Collins, MCSE
Systems Manager
Nesbitt Engineering, Inc.