DHCPv6 PD server
phil at philkern.de
Thu Aug 2 15:07:35 CEST 2012
On Thu, Aug 02, 2012 at 02:56:35PM +0200, Seth Mos wrote:
> I sent a message on the ISC dhcpd list but got no response. My most
> educated guess is that they do not have the resource to implement a
> daemon that inserts routes into the routing table yet.
but they have it on a list of known deficiencies?
> Note that this also applies to the dhcp6 relay. Which would need to
> insert routes.
True enough. This requires more logic than just en-/decapsulate those
DHCPv6 queries/answers. And it's also hard to do that from the outside
if it's not supported by the device, as I guess there's no proper
solution to inject routes into the routing table for things that are
not local to the injector. (Like through a routing protocol or using SNMP from
> Because ISC dhcpcd is running under a unpriviledged user a 2nd
> daemon would be required/signaled for the route insert/remove/update
I presume you mean dhcpd, not dhcpcd? Well, given recent developments
in the Linux world I wouldn't be surprised about a dbus call trying to add
routes and somebody else picking them up. ;-)
In this case a proper signal/hook would already be "enough" for a first stab,
> I briefly looked at other servers but had pretty much the same
> conclusion. This works because the leasefile is written when a new
> lease is handed out, this also causes you lease per second
> performance of your DHCP server to be limited by disk IO.
True. For a recent conference I had to run the DHCPv4 server with eatmydata
because it actually fsync()ed for every lease, which was horrible for
desktop hardware. (There seems to be a configure option for this, too.)
I guess events for the modified file would fire in the same instant even if the
content does not reach persistent storage space.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 588 bytes
Desc: Digital signature
Url : http://lists.cluenet.de/pipermail/ipv6-ops/attachments/20120802/89a0311c/attachment.bin
More information about the ipv6-ops