DHCP service fails to restart correctly after power cut

Question

Level 1

85 points

DHCP service fails to restart correctly after power cut

We've had a couple of major (> 7 hours) powercuts, which cause our servers to shut down once the UPS is drained.

They do restart once power is restored, but the DHCP service has on two occassions failed to start correctly. Nothing is visible in the logs, and I'm tempted to just stop DHCP service on the Tiger Server, and run it instead on a very stable firewall appliance we have (pfsense box on ALIX hardware FWIW).

The only issue with my proposed solution is that the DHCP server on the firewall appliance would need to know the custom client information to supply so that MacOSX client machines get the LDAP server address correctly, needed for authentication and auto-configuration.

I suspect the custom client information I need to specify on the DHCP Server is:

ldap_url="ldap://MY.TIGERSERVER.ADDRESS/dc=MYDOMAIN,dc=com"

Is this the case? Or does anyone have any pointers on why the Tiger Server is not starting DHCP correctly?

The tiger server is configured to also serve DNS to LAN clients.

TIA,

Mezza

XServe G5, Mac OS X (10.4.11)

Posted on Dec 21, 2009 4:03 AM

Reply

Answer 1

Camelot

Level 9

58,581 points

Dec 21, 2009 11:48 AM in response to Mezza

It's not clear whether DHCP doesn't restart after a power failure, or if it doesn't restart after any reboot (intentional or otherwise). You only mention 'two occassions' - are they both power-related? or was one a scheduled reboot?

In addition, you say it doesn't 'restart correctly', which implies it restarts at some level...?

Then, of course, there's the question of what you do to get it working 'correctly'? What actions are you taking? That might provide some clue, too.

Reply

Answer 2

Mezza Author

Level 1

85 points

Dec 21, 2009 12:11 PM in response to Camelot

Perhaps I could be clearer.

The issue is that, on the two recent occasions that the server has restarted from a power related shut down, the DHCP service has failed to assign IP addresses to any LAN devices.

ServerAdmin (both GUI and CLI) reports that the service is running, but trying to boot a LAN machine results in no address being assigned to the client.

The resolution in both cases has been to manually restart DHCP from the CLI.

I've tailed the system.log during restarts of the DHCP service but nothing meaningful is shown.

Reply

Answer 3

Dec 21, 2009 12:15 PM in response to Mezza

Hi

The custom information is not really necessary to be honest. It's only useful if you want to provide all of the information for your environment to client workstations via DHCP. All that really matters is which server client workstations are using to resolve their DNS. You could assign that manually or - better still - via DHCP.

In an OD environment and if you move DHCP to your hardware appliance then make sure the primary DNS Server IP 'assigned' to your workstations is the OD Master.

In addition to Camelot's advice I would begin troubleshooting the Server's drive/directory structure. A powercut is never a good thing. If the Server's Boot drive is not mirrored (ie: only a single drive) then there is that risk. What does DU say? Have you tried fsck from the command line yet? Do you have a copy of DiskWarrior handy? Make sure you have a recent and effective backup. I would archive the LDAP Database if I was you whilst you still have a chance. Even if there ends up being no problems It's still a good idea.

Tony

Reply

Answer 4

MrHoffman

Level 10

146,563 points

Dec 21, 2009 12:25 PM in response to Mezza

Do other protocols have connectivity, or is this just DHCP that's adversely affected here?

I've seen similar bad behavior when the network hasn't stabilized before the clients and servers get rolling, and when the network itself is blocking the DHCP multicast traffic, and when a rogue DHCP server is jamming the network.

Crank up the logging settings on the dhcp daemon.

Use Wireshark or other such to have a look at the network traffic, too.

Also confirm the local IP network is correctly configured (changeip -checkhostname) and that the forward and reverse DNS settings are correct.

How many NICs are present here? One? Or more than one?

Reply