mike
Aaron Gaudio wrote:
Behold, Mike Cooper <Mike.Cooper@xxxxxxxxxxx> hath decreed:
I'm having trouble with FC2 NIS clients (all are X86/IA32) not being able to bind/find a NIS server if the FC2 client is set to "broadcast" via
domain reshape.com broadcast
in /etc/yp.conf
Apparantly ypbind never finds a server, even though there are 2 on the same subnet. I ran ypbind -debug and it seems to get RPC timeouts when it broadcasts for a server. A "ypwhich" command fails after about 90 seconds since ypbind eventually exists once it can't find a server.
My RHL 7.2 - 9 clients have no trouble with the same yp.conf config.
The NIS servers are both Solaris 8 with Sun's stock NIS ypserv.
If I change yp.conf to be
ypserver 10.X.X.X1 ypserver 10.X.X.X2
It does bind, but *very* slowly. NIS lookup is extremely slow. If not for nscd, it would make the whole system unusably slow.
And yes, I know that "broadcast" isn't the safest thing in the world, but it's right for this environment.
Anybody have any clues on this?
I don't know if this explains your situation any, but after upgrading here at work, ypbind was never connecting. I also use broadcast, but the problem was not ypbind, it was the fact that I had two ethernet cards and the system was loading the module for (what should be) eth1 before eth0, causing the device names to get swapped. Strangely, if I booted into single user mode and manually executed 'service network start', they got loaded in the correct order.
My solution was to add the following to /etc/modprobe.conf:
nstall ne2k-pci /sbin/modprobe eth0; /sbin/modprobe --ignore-install ne2k-pci
(note ne2k-pci is what is supposed to be eth1).
This ensures that if something causes ne2k-pci to get loaded, it makes sure to first load eth0 (which is an e100 card).
