How to run in failover but NOT load balancing?

Glenn Satchell Glenn.Satchell at uniq.com.au
Sat Dec 22 12:47:29 UTC 2007


You could use failover in it's usual load-balancing way, and use a
separate script to automatically switch to failover after a
pre-determined amount of time. In normal use, where the load is split
evenly between two failover peers, there is no immediate outage if one
peer goes down. If you search the archives there are several
suggestions for switching to partner down mode using omshell.

One of the HA solutions may also be suitable for you. It really depends
on why you feel you only need one server active. Generally that means a
simple dhcp configuration, but a complex OS or HA configuration.

regards,
-glenn

>From: "Benjamin Wiechman" <benw at meltel.com>
>To: <dhcp-users at isc.org>
>Subject: RE: How to run in failover but NOT load balancing?
>Date: Fri, 21 Dec 2007 16:27:58 -0600
>
>If you want a solution like that would it not maybe be better to look at
>Linux HA and DRBD or some similar mechanism. I know MySQL touts this as a HA
>possibility for their DB. There is no reason it shouldn't work for DHCP. 
>
>I would guess that a HA based solution would provide better QOS in the long
>run. Using the split directive to allocate all IPs to one failover peer lead
>you to a condition where no leases would be assigned if the peer responsible
>for handing out the leases went down. You would basically have to duplicate
>the functionality of the HA package to detect that outage on the backup peer
>and put that server in a partner down state, or do it manually - which may
>lead to a much longer outage than using HA, or very similar results.
>
>Ben Wiechman
>Network Admin
>Wisper High Speed Internet
>ben at wisper-wireless.com
> 
>> -----Original Message-----
>> From: dhcp-users-bounce at isc.org [mailto:dhcp-users-bounce at isc.org] On
>> Behalf Of Stulic,Damjan
>> Sent: Friday, December 21, 2007 8:15 AM
>> To: dhcp-users at isc.org
>> Subject: RE: How to run in failover but NOT load balancing?
>> 
>> set split to 0 or 255 to turn off load balancing.
>> but be aware (from my experiences):
>>  - both servers will still log the dora process, and keep leases in
>> dhcpd.leases file
>>  - on true failover (when primary is down) some functionality is
>> unavailable (like to add new pools)
>> 
>> Damjan Stulic
>> IS Security Identity Management
>> Edward Jones
>> 
>> 
>> 
>> 
>>  If you are not the intended recipient of this message (including
>> attachments), or if you have received this message in error, immediately
>> notify us and delete it and any attachments.  If you no longer wish to
>> receive e-mail from Edward Jones, please send this request to
>> messages at edwardjones.com.  You must include the e-mail address that you
>> wish not to receive e-mail communications.  For important additional
>> information related to this e-mail, visit
>> www.edwardjones.com/US_email_disclosure
>> 
>> ________________________________
>> 
>> 
>> 
>> From: dhcp-users-bounce at isc.org [mailto:dhcp-users-bounce at isc.org] On
>> Behalf Of Nomad
>> Sent: Thursday, December 20, 2007 18:56
>> To: dhcp-users at isc.org
>> Subject: How to run in failover but NOT load balancing?
>> 
>> 
>> I think I've misunderstood something in setting up DHCP 3.1.0.  I have
>> two servers and I want them to run in failover mode, not load-balancing,
>> with one server providing service to all dhcp requests and the other
>> acting as a standby.  I have both listed in the router's configuration
>> (ip helper-address).  Is this supported under the failover protocol
>> implemenation?  Or is the only way to run two servers to have them load
>> balance?  I'd like to run in the active-standy mode to simplify
>> troubleshooting so I don't have to figure out which server has acted on
>> a client's request.  Initially I thought I could do this with the
>> "split" statement but everything I've read tells me I should keep the
>> split value at  128  which implies balancing between the two servers.
>
>
>
>


More information about the dhcp-users mailing list