Migration of leases from one failover pair to another without outage
fabric at northwestern.edu
Wed Feb 27 15:17:45 UTC 2008
On 2/27/08 12:03 AM, Nick Urbanik wrote:
> Or indeed, I'd love to hear any other suggestions for migrating leases
> from one failover pair to another without an outage.
How about this, it's a little daring, and you'll need to up your lease
time probably somewhat significantly, so I you can't to that, it's
probably not an option. Nor if you need the lease time to be short again
Increase the lease time to something fairly significant. I'm think like
a week, but a couple days might suffice. The goal here is to get the
leases file to be essentially static for the duration of the switchover.
By making the leases file so long, you should nearly eliminate the
possibility of clients doing a DHCPREQUEST at half lease time.
Unfortunately you'll probably cause problems for new client requests, so
you'll want to make the change when the least number of new clients
are coming online.
Once you've established this very low turnover leases file, you can
simply shut down secondary a, and then transfer the leases file from
secondary a to secondary b, as well as update the config. REmove
secondary A from the helper addresses. Startup secondary b. I'm
assuming the IPs on secondary a and be are different, in which cases
they won't negotiate the failover peering, but should both be able to
answer queries. Oonce secondary B is started up, you can update the
helper addresses to be only secondary B, and shutdown primary A. Now
update primary A's config to point at peer secondary B. Start it up,
and once it's up and running, re-add the helper address for primary A.
That's pretty convoluted, but it should really minimize downtime, and
only impact a small number of clients. I"m not sure you can do this
with 0 downtime and no client issues, but the few clients should recover
fairly quickly, but there could possibly be some fallout.
If you just do an IP swap of secondary a and secondary b, I bet this
could be done much more easily. Just shutdown secondary a and remove
it's IP from the helper addresses. Swap IPs, restart the server that
was formerly secondary b, and let DHCP sync up the files. Primary A
should continute to answer queries in a peer-down state. Once the now
seconday A box is back online, add the helper addresses back in.
That's probably what I'd do. chris
More information about the dhcp-users