I also believe there is information in the wiki or release notes about workarounds.
I have had a somewhat similar problem with some servers at work. The
servers in question acquire their ip address from a dhcp server and
either mount some filesystems over nfs or, in case of the nfs server
itself, export the relevant directories to specific hosts. During
bootup, dhclient tries to acquire an ip address but fails and
backgrounds since the network link is not yet activated. The link is
activated some time later in the boot sequence and dhclient succeeds on
its second attempt, the default being after several minutes of the first
attempt. However it is too late for the system, since by that time both
the nfs mounting and the nfs server process would have failed. The
server process fails because it is not able to resolve the host
addresses to which it is supposed to allow access.
I had realized it was some timing issue with the network driver. I tried
explicitly loading the network related modules at the
S20module-init-tools stage by adding the module to /etc/modules. I have
added a couple of echo statements to the modules-init-tools in order to
verify that the network module is indeed loaded before the networking
stage. However the problem still persisted.
I was finally able to workaround the problem by adding a 60 second sleep
at the end of /etc/init.d/networking and reducing the dhclient timeout
to 5 seconds and the retry to 10 seconds. This way, dhclient backgrounds
some 5-10 seconds after starting, but is able to succeed the second time
before the end of sleep.
I wonder if there is some better way to do it though. The link provided
by Marc describes this issue, but I could not find any workaround anywhere.
The relevant network module in my case was: bnx2
Thanks & regards,
Raj Kiran Grandhi
To UNSUBSCRIBE, email to debian-user-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact email@example.com