Red Hat Bugzilla – Bug 198608
DHCP very slow
Last modified: 2007-11-30 17:11:37 EST
Description of problem:
anaconda took over 90 seconds to get a DHCP address on this box during
kickstart; it was trying various DHCPv6 things in the mean time, it seems, which
were repeatedly timing out.
14:24:10 INFO : only have one network device: eth0
14:24:10 INFO : sending dhcp request through device eth0
14:25:47 ERROR : DHCPv6 eth0 - TIMED OUT - failure.
14:25:48 INFO : DHCPv4 eth0 - obtained lease 172.16.18.103
(the subnet this box is on gets automatically assigned IPv6 addresses via radvd
if that makes a difference)
Created attachment 132291 [details]
That's normal. You need to set a time-out to see if the "DHCP" request
is honored or not. Add "noipv6" to the boot options or uncheck "ipv6"
support in the installer network configuration to skip this step.
Both DHCPv4 and DHCPv6 time out after 45 seconds. But that's 45 seconds for
each protocol, hence the 90 second total.
There seem to be some other issues regarding radvd IPv6 configurations and the
anaconda loader right now. Looking in to that.
To clarify: the DHCPv4 request did not time out at all, it succeeded
immediately, as the logs show; so that should not account for half the 90
seconds of timeout.
I'm not sure if it's worth continuing to wait for a DHCPv6 reply once the
interface gets assigned an address from the radvd broadcast; perhaps the timeout
could be decreased at least in that case.
For RFC compliance, we stick with 45 seconds for the DHCP client request. But
special handling for radvd needs to be present so we then don't run the DHCP client.
Marking this report closed, notabug.