problem with network initialization in SuSE 10???
Hi, Okay folks! Please help, this is driving me nuts. Apologies for the X-posting. I have a AMD64 machine (SuSE 10 64bit) connected to a switch. At boot time my PC tries to NFS mount 3 exported partitions from a SLES server attached to the same switch. The PC fails to mount any of the NFS shares at boot time. I suspect that there could be a couple of things broken, since if I insert a ping command at the end of the start) section of etc/init.d/network, my machine reports "no route to host". Leading me to think that the network really isn't up. So, I inserted a `sleep 20` prior to the ping command, the machine sits there, then reports "no route to host". Eventually, the network must come up, since if I then log in through KDM I can browse the web etc. But the only way I _ever_ get my NFS shares mounted is if I issue a `mount -vat nfs` command manually. Now I don't mind waiting for the network to come up, but shouldn't SuSE be waiting too? I.e. what's the point in trying to mount NFS shares if you don't have a working network? FYI we have several x86 machines and a single AMD64 (SuSE 9.3 64bit) that start up and NFS with the server with no problems. As always, any help much appreciated. Cheers, Jon.
On Monday 06 February 2006 13:43, Jonathan Brooks wrote:
I suspect that there could be a couple of things broken, since if I insert a ping command at the end of the start) section of etc/init.d/network, my machine reports "no route to host". Leading me to think that the network really isn't up.
It's probably some problem talking to the switch. Double check the cabling and the switch configuration.
But the only way I _ever_ get my NFS shares mounted is if I issue a `mount -vat nfs` command manually.
Now I don't mind waiting for the network to come up, but shouldn't SuSE be waiting too? I.e. what's the point in trying to mount NFS shares if you don't have a working network?
You could file an enhancement bug in bugzilla.novell.com. But i wouldn't hold my breadth of that it gets fixed because it's a nasty problem and making things synchronous at boot time has a couple of drawbacks (for once it's slower). Better you just use autofs instead. In general putting NFS mounts into fstabs is fragile and not really good practice and automounters are designed to deal with all these problems. I remember long ago seeing a cluster of machines that had circular dependencies in their boot time NFS mounts. Generated quite a mess when the cluster had to be rebooted. -Andi
Andi Kleen wrote:
I remember long ago seeing a cluster of machines that had circular dependencies in their boot time NFS mounts. Generated quite a mess when the cluster had to be rebooted.
A long time ago when I worked in a support department, I used to make sure the machines were configured like this. It stopped anybody stealing one of the machines to replace a failed customer machine :) And yes, getting the boot scripts right was a pain! Cheers, Dave
Never ever use PLANET switches, they suck. I pulled the patch lead from the crummy PLANET switch and "switched" to a 3com one - now everything works absolutely fine. Thanks for your suggestion Andy. Cheers, Jon. Andi Kleen wrote:
On Monday 06 February 2006 13:43, Jonathan Brooks wrote:
I suspect that there could be a couple of things broken, since if I insert a ping command at the end of the start) section of etc/init.d/network, my machine reports "no route to host". Leading me to think that the network really isn't up.
It's probably some problem talking to the switch. Double check the cabling and the switch configuration.
But the only way I _ever_ get my NFS shares mounted is if I issue a `mount -vat nfs` command manually.
Now I don't mind waiting for the network to come up, but shouldn't SuSE be waiting too? I.e. what's the point in trying to mount NFS shares if you don't have a working network?
You could file an enhancement bug in bugzilla.novell.com. But i wouldn't hold my breadth of that it gets fixed because it's a nasty problem and making things synchronous at boot time has a couple of drawbacks (for once it's slower).
Better you just use autofs instead. In general putting NFS mounts into fstabs is fragile and not really good practice and automounters are designed to deal with all these problems.
I remember long ago seeing a cluster of machines that had circular dependencies in their boot time NFS mounts. Generated quite a mess when the cluster had to be rebooted.
-Andi
-- Jonathan Brooks (Ph.D.) Research Assistant. PaIN Group, Department of Human Anatomy & Genetics, University of Oxford, South Parks Road, Oxford, OX1 3QX tel: +44(0)1865-282654 fax: +44(0)1865-282656 web: http://www.fmrib.ox.ac.uk/~jon
participants (3)
-
Andi Kleen -
Dave Howorth -
Jonathan Brooks