Issues with next-server & filename

Matt Simmons standalone.sysadmin at gmail.com
Tue Nov 30 21:21:32 UTC 2010


Hi All,

I just joined this list, and while I scanned some of the archives, if
I get the etiquette wrong, please forgive me.

I've been having some significantly weird issues with dhcpd for the
past couple of days. Here's the brief overview:

At one of my sites, I have two VM guests on the same VM host, one of
them being a PXEboot server, the other being a client machine which
needs re-installed often. The dhcpd host is running CentOS 5.5,
completely updated. The version of dhcpd is isc-dhcpd-V3.0.5-RedHat.
This machine works perfectly.

At another site, I needed to mirror that configuration, so I did, as
closely as possible. Two new VM machines were built, a dhcpd host was
installed and configured from the same DVD ISO, both have the same
packages, both are completely up to date, and the dhcpd configuration
is mirrored, with the exception of the netblocks and the MAC address
of the client. This machine does not work perfectly.

After capturing the packet stream, I have determined that the only
difference in the DHCP OFFER packet is that the working machine
specifies the "Next server IP address" and "Boot file name", while the
broken one does not.

Here are the two configurations:

Working:
      1 ddns-update-style interim;
      2
      3 subnet 10.11.1.0 netmask 255.255.255.0 {
      4
      5  option routers    10.11.1.1;
      6  option subnet-mask   255.255.255.0;
      7  option domain-name   "mydomain";
      8  option domain-name-servers 10.11.1.43;
      9  option time-offset   -18000;
     10  range dynamic-bootp  10.11.1.95 10.11.1.96;
     11  default-lease-time   21600;
     12  max-lease-time    43200;
     13
     14  group {
     15     next-server 10.11.1.91;
     16     filename "pxelinux.0";
     17
     18     host ops1tt {
     19        hardware ethernet 00:0c:29:05:6a:82;
     20        fixed-address  10.11.1.94;
     21     }
     22  }
     23 }


Not working:
  1 ddns-update-style interim;
  2
  3 subnet 10.10.1.0 netmask 255.255.255.0 {
  4
  5    option routers          10.10.1.1;
  6    option subnet-mask      255.255.255.0;
  7    option domain-name      "mydomain";
  8    option domain-name-servers      10.10.1.43;
  9    option time-offset      -18000;
 10    range dynamic-bootp     10.10.1.95 10.10.1.96;
 11    default-lease-time      21600;
 12    max-lease-time          43200;
 13
 14    group {
 15       next-server 10.10.1.91;
 16       filename "pxelinux.0";
 17
 18       host ops1tp {
 19          hardware ethernet  00:0c:29:2d:ea:5a;
 20          fixed-address   10.10.1.94;
 21       }
 22    }
 23 }

Since there's no effective difference in the configurations that I can
determine, I tried to rule out environmental considerations. In
addition to checking the software packages installed on each machine
(identical), I verified that SELinux is in permissive in both and that
the VMware hosts were identical (yes, ESXi version 4.1.0, 260247).

I have removed and reinstalled the dhcpd package using yum (same MD5
checksum each time, identical to the one that works), and in a moment
of desperation, I even created another virtual machine configured
identically, and it shows exactly the same results.

I am very near to becoming a gnostic deist based on the plain evidence
that there must be a God, with the corollary that he hates me.

Can anyone help me shed some light on this? Logically, there HAS to be
something up with the config, right? Since nothing is changing but the
numbers, is there something magical about the 10.10.1? I don't know
what else to do at this point.

Thanks very much,

--Matt




-- 
LITTLE GIRL: But which cookie will you eat FIRST?
COOKIE MONSTER: Me think you have misconception of cookie-eating process.



More information about the dhcp-users mailing list