Description of problem:
Same bug as http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=590641
We have a server with one NIC, on two logical networks, and DHCP won't work with the secondary (alias) interface and subnet. Is there a fix for this? This worked in DHCP v3, but not in v4.
Version-Release number of selected component (if applicable):
Here is the exact error:
No subnet declaration for br0:1 (no IPv4 addresses).
** Ignoring requests on br0:1. If this is not what you want, please write a subnet declaration in your dhcpd.conf file for the network segment to which interface br0:1 is attached. **
... except that I *do* have a subnet declared that is on (and only on) br0:1 !
Yes, this is a "common bug" in dhcp-4, see bug #449946 and
There is a patch, however it's quite invasive and I'd rather not diverge from upstream *so much* (I don't see it much likely that they ever merge it).
I'm sure this bug bothers you but I'd rather not use this patch in RHEL-6, because it could break a lot of other stuff.
What about this plan ? :
- I'll make a RHEL-6 scratch build (testing build that won't make it into official repositories) for you with this patch applied.
I just need to know which architecture (x86_64/i686 etc.) do you use so I can send proper testing packages to you.
- If you confirm that it fixes the problem for you and you see no regressions I'll finally fix bug #449946 in Fedora so the patch gets some more testing in Fedora and will eventually make it into RHEL-7.
I'm closing (however we can still discuss it here) this ticket as NEXTRELEASE because it could eventually be fixed in RHEL-7.
But it won't be fixed in RHEL-6 from reasons stated earlier. Thank you for understanding.
Awesome! RHEL 6.2 x86_64 is what we are using.
I can test said scratch packages for you immediately
Created attachment 573612 [details]
Use getifaddrs() for interface discovery code on Linux
(In reply to comment #3)
> Yes, this is a "common bug" in dhcp-4, see bug #449946 and
> There is a patch, however it's quite invasive and I'd rather not diverge from
> upstream *so much* (I don't see it much likely that they ever merge it).
As I stated in bug #449946, comment #9 I had figured out another solution than the patch from dhcp-hackers mailing list. It's simple - it actually throws out all the interface discovery code (380 lines) for Linux and uses the same code as for *BSD/OSX, i.e. getifaddrs() function.
Therefore it shouldn't break anything and also could be acceptable for upstream.
> But it won't be fixed in RHEL-6 from reasons stated earlier.
I changed my mind. Reopenning for RHEL-6.
Technical note added. If any revisions are required, please edit the "Technical Notes" field
accordingly. All revisions will be proofread by the Engineering Content Services team.
Network interface of a server has 2 or more configured IP addresses and the DHCP server itself is configured to serve subnets to whose the second (or any next, not just the first) IP address belongs.
DHCP server is able to discover only the first IP address of network interface and therefore is able to serve only subnets to which only the first address belongs. If the server is configured to serve only subnet to which the second or any next IP address of the network interface belong, it refuses to start with 'No subnet declaration for ...' message.
Network interface addresses discovery code was completely changed.
DHCP server is now able to find all addresses of a network interface and therefore is able to serve subnets to which these adressess belong.
Verified on RHEL6, dhcp-4.1.1-34.P1.el6.x86_64. dhcpd is now able to start and provide service when configured to bind to secondary IP address only.
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.