[ednog] Anyone using the ISC DHCPD failover protocol?
aikasevj at cac.washington.edu
Tue Sep 20 17:50:20 EDT 2005
On Tue, Sep 20, 2005 at 05:28:53PM -0400, Andrew W. Elble wrote:
> Hi all!
> Quick question, is anyone using ISC's DHCP server configured with
> the failover protocol/replication enabled? What have your experiences
> been? Would you recommend it in a production environment?
We here at the University of Washington have been using it for quite a while
now. We run it in the failover mode (4 servers in clusters of 2) and have
taken couple of failures that have caused a single machine outage (disk,
etc.) that have allowed us to keep on serving leases from the other server.
Sometimes you have to do little bit of tweaking to bring the failed machine
back to the system (can't remember the particulars right now), but overall
been great. Also, when bringing machines down for maintenance, there's a
'right' way to do it, which can save lots of headaches when bringing the
machine up again. The documentation is fairly good at explaining all this
We also had to patch the code (while back though) to get the failover
protocol to work on redhat linux. Last time we looked the patch was not
We also use OMAPI interface to muck with the lease states (for network
access control), and this has caused us little bit of grief in the past as
the lease databases on the different servers can get out of sync. Also of
course you have to do OMAPI calls to both places as this is not synced
between the two servers.
We tested the failover protocol and performance/scaling pretty extensively
prior to deploying, and wrote some test programs
(http://staff.washington.edu/aikasevj/dhcp_client/) that might be of
interest prior to deploying.
So, overall, it's definitely helped in a face of failure :) If you have any
specific questions, or need to bounce something off, feel free to contact
either on or off the list.
> ednog mailing list
> ednog at puck.nether.net
More information about the ednog