MS DNS and Bind on Linux play well togther, but does anyone know how to do the same for DHCPD? I'll avoid the political bullshit story here, but basicially I need a way to have a DHCP server on Windows and a backup on a Linux box.

I'm not familiar with a data synchronization utility, but dual well-behaved DHCP servers on a network should be fine as long as the data between them is the same.

but dual well-behaved DHCP servers on a network should be fine

I know the Linux DHCPD does ping an address before it is assigned to ensure it is not in use. So, 2 running the same software should be fine. But I can't find proof anywhere that the Microsoft DHCP server does the same, as the MS guy at work is really paranoid about running 2 DHCP servers in the same scope.

Just found this technet article stating at least the MS DHCP server can be configured to ping an address before handing it out. I'll have to check the MS box to see how it is set up currently, I didn't remember stumbling across this option. (GUIs easier to use on a server, yeah right.)

Still haven't found anything about sharing lease files, but this should be good enough. Oh, to bring up part of the political reason...

Anyone here an HP customer with Proliant servers? If you called in to tech support, and asked the guy what your servers were running, and his answer was "A mixture of Windows and Linux", what would your response be? Role play this a little as a Windows admin. And yes, this was a reason I was given by a manager to not outright replacing a virus ridden piece of crap Windows 2000 server that the team uses, even thouh the majority of the team was in favor of such an action. He seems to forget that HP actually has a policy not to connect an Exchange server directly to the internet, and that any external mail must pass through a Unix box somewhere. He also seems to forget we make 2 versions of Unix ourselves (Tru64 and HP-UX), plus VMS, and used to have Bruse Perens on the payroll.

Grr...

Easy solution;

Set up windows server.
Set up linux server.
Shut down windows server and give the motherboard a good zapping with a defibrilator.
Pretend not to notice for a few weeks, and then use the linux server's success as a soap box.


I still can't comprehend why someone would want to have to administer 2 different implementations of dhcpd on 2 different OSs (or pay for said administation), and try to keep them in sync. Surely in todays environment, security is an integral part of what a dhcp server does, and having 2 server opens more opportunities for attack. Would you notice if a 3rd server suddenly appeared and started handing out IPs?

One way that you could make this work a little easier for you would be if you locked down client MAC address to host name maps on the main server, and only supply the locked down IPs from that server, and on the other server only supply generic IPs from a different range. The generic IP range could have a more restrictive policy in effect, so effectively they can only access enough to notify a sysadmin.

That means that your trusted machines should always pick up IPs from the main server - if they all start getting resticted IPs then you'll know quickly that the main server has stopped handing out IPs for some reason.

Of course, I'd still want the linux box as primary.

One way that you could make this work a little easier for you would be if you locked down client MAC address to host name maps on the main server, and only supply the locked down IPs from that server, and on the other server only supply generic IPs from a different range. The generic IP range could have a more restrictive policy in effect, so effectively they can only access enough to notify a sysadmin.

I've inherited a very similar setup, and this is essentially the method we use. Both servers have the same MAC->Static IP table, and either one can respond to a given request on the network. They also have a pool of "Roaming" IPs, with one server handing out odds, and the other handing out evens.

Matthew

Shut down windows server and give the motherboard a good zapping with a defibrilator.
Pretend not to notice for a few weeks, and then use the linux server's success as a soap box.

As if you'd even need to sabotage the Windows server? Heck, just let it run, it'll go belly up all on its own.

As if you'd even need to sabotage the Windows server? Heck, just let it run, it'll go belly up all on its own.

It already has, twice now. Once requiring a repair using the 2K CD, and the second time all the partitons were gone.

I'd love just to power the Windows box off, but I'm only the backup admin for it, meaning the main guy has been the one rebuilding it, and supporting it. I am the admin of the Linux server, that has worked fine while he tinkers with the 2000 box.

Oh, this guy flat out told me "We live in a Windows world". His background? Desktop tech support. At least the old Windows admin wasn't an ass about everything. Should have made a move to switch everything over back then, would have had less resistance to it.

Anyone here an HP customer with Proliant servers?

Why that would be me. And ours are purple.

If you called in to tech support, and asked the guy what your servers were running, and his answer was "A mixture of Windows and Linux", what would your response be?

I'd say you have too many Windows machines. Actually, my gripe with HP and OS's is that even though you're now distributing Mandrake on the desktop machines, the Proliant guys still can't figure out that Redhat != Linux. Just try to get the health monitor or Insight agent running under Debian.

the Proliant guys still can't figure out that Redhat != Linux. Just try to get the health monitor or Insight agent running under Debian.

Actually, they do support more then RedHat. Support also exists for SuSE, TurboLinux(big in Japan) and the UnitedLinux based distros.

Of course they are all RPM based, so I can see how Debian might not take kindly to that. Trust me, many of us on the inside beat the developers up over Linux interopibility issues quite a bit. And they have gotten better. The initial agents would require specific kernels, and were slow to keep up with errata kernels. That has since been solved in a similar way that NVidia does their Linux drivers.

I really wish we hadn't lost Bruce Perens. I'm sure if he was still at HP, there would be more Debian support.

edit: Oh, that reminds me. Have you upgraded to the newest firmware for the Remote Insight boards (and integrated varaties)? It has official support for Mozilla under Linux, including I believe working virtual CD-ROM on the RILOE II.

Of course they are all RPM based, so I can see how Debian might not take kindly to that.

I admit that I haven't tried the obvious lately:

apt-get install rpm
rpm -Uvh hpasm....rpm

That prompted a descent into kernel and libc version hell last I tried it, but it might work OK now. I think I'm current on firmware, but all of our machines have iLo's so I don't think the Virtual CD works (that's RILOE II only, right?)

I think I'm current on firmware, but all of our machines have iLo's so I don't think the Virtual CD works (that's RILOE II only, right?)

Ahh, we don't quite have all the firmware releases out. I think the iLo one will be 1.40, quite a few changes in it. To give you an idea, here is the RILOE II update list. The iLo list will have a few more things:

* Use directory services for user authentication and authorization.
* Enhanced virtual floppy drive operation.
* Support for LINUX client browsers.
* Performance improvements to remote console.