I don’t have a magic bullet answer. I’ll suggest some avenues of attack. Wireshark or tcpdump should help on the timeout. Check the firewall logs or briefly turn off the firewall. Also ensure you are not mixing nfs3 and nfs4. Make sure the nfs4 root has an fsid of 0. Are you using kernel nfs server? I’ve had the best luck with that.
I’m quite sure that the local network won’t be attacks target, the firewall (I have poor knowledge in firewall and iptables) is set on both computers but set in internal zone because if not, I couldn’t use samba, nore nfs, nore any sharing facilities even if “add port in firewall” is set !!
I use an ADSL router that is not forwarding ports so (I did the experience) the port 80 is not accessible from the external dangerous jungelous area ^^ (for example) I can’t join the internal web server running in the LAN by any computer on a different network (from work for example). I got no public IP, and even if I try with the IP provided to my ADSL router I can’t pass thru (I will dig to make it possible because before I switched to ADSL I used a cable modem and it was working better).
When I check for mount -t nfs 192.168.1.2:/home/AnEntryToMount /media/nfsmounted -rw --verbose it indicates nfs3 on both servers (the one in 11.1 and the one in 11.3)
I use nfs-kernel-server on both computers, only the version changed.
One thing I found strange, on the 11.3 server, when I go to software management and type nfs-client I saw that the actual version was 1.2.1-7.2 and there was a 1.2.1-8.3.1-586 from OpenSUSE-11.3-Update, I did the update but it seems there was no effect… Yast displayed errors (no time to read nore redirect to check them)…
The issue is not scaring because I can mount without any problem and access as root/user but there is different behaviour from one version to another that’s all, I just have to specify the host when doing showmount -e (maybe because there is two nfs servers on the same network ?? I did a check with the one in 11.1 off, with no exports entries and nfsserver set to off by /etc/init.d/nfsserver stop), but it changed nothing…
Maybe once again this is a strange behaviour because I still have not change my habits to run applications with a uid different than 0 ?! (I used root accoutn since 1996 and crashed my Linux twice within 14 years and again it was the kde environnement I did recover all my system by restoring the kde environnement).
I don’t tried it using another user… I’m checking… ok, showmount will not be used as a “simple” user with restricted permissions/privileges/rights/…
For your Server and client machines.
did you enable portmap in Yast → System → System Services (runlevel)?
On your servers, in the NFS setup in Yast have you checked “Enable NFSv4” and put in a domain name? (The default, localdomain, should be fine). Make sure both servers have the “enable NFSv4” option checked, and the domain name is the same.
let us know!
When I type rpcinfo on both computers I got similar results.
I got all kind of stuff that is quite similar from 11.3 machine to 11.1.
portmap works otherwise it could be impossible to mount volumes from 11.3 or 11.1 but I can mount volumes from both “side” of the network.
Make sure both servers have the “enable NFSv4” option checked, and the domain name is the same
I unchecked on both computers…
When I check for mount -t nfs 192.168.1.2:/home/AnEntryToMount /media/nfsmounted -rw --verbose it indicates nfs3 on both servers (the one in 11.1 and the one in 11.3)
I use nfs-kernel-server on both computers, only the version changed.
This anwered the advice :
Also ensure you are not mixing nfs3 and nfs4.
I’m a little lost… maybe the kernel versions that are different may be the point, the libraries available are surely different from one kernel to another… and the behaviour will follow…
I just found a similiar problem with our SLES 11 SP1 showmount (nfs-client-1.2.1-2.24.1).
This version of showmount isn’t using the system default tcp anymore for it’s rpc query but tries to use UDP for this.
As we open (external) firewalls in this case only for the three basic nfs tcp ports (so without udp) showmount is not working from that mashines anymore but a mount succeeds as the tcp protocol (or the one specified in opptions) is used there.
rpcinfo -p also works fine (uses TCP per default or some kind of fall-back machanism what is what I would expect if it cannot be configured explicitly).
So someone should propably check the sources for SuSE’s shownmount in recent release 11 version.
On 05/03/2013 05:36 PM, EliasOhm wrote:
>
> I just found a similiar problem with our SLES 11 SP1 showmount
> (nfs-client-1.2.1-2.24.1).
> So someone should propably check the sources for SuSE’s shownmount in
> recent release 11 version.
you have (inadvertently i guess) posted to the openSUSE forums, the
SLES forums are over at http://forums.suse.com/
the ID/Pass used here will work there…
there are precious few here with any SLES experience, and pratically
zero cross feed…would probably be best to write a bugzilla, or at
the very least contact the SUSE Linux Enterprise folks.
–
dd
openSUSE®, the “German Engineered Automobile” of operating systems!
No, it was not inadvertly
I was searching for a solution for the problem with showmount because a collegue of the linux team asked me why it is not working from a SLES 11.1 mashine, as he is used to do allways a showmount befor doing the actual mount he previously reported he cannot access the the share, which is not true, for default mount command on SuSE Linux >= 9.
And I found this article on google, first entry And it is regarding the same package version. I do not know the SuSE firewall more deeply so I cannot say whether the same problem does apply here. But it sounds like that!
You are probably rigth, I can post at SLES Forum. But I did not found that on google (at least not in time, with my search query) and I’m not in the linux admin team of our company, otherwise I would have done that first, of course. (I do not have a company account for Novell as well, if that matters?)
I could even post a patch that propably works for both versions, SLES and OpenSUSE. I would only need to take some time for that. I could make showmount using tcp first and than udp if it fails (at least for the rpc query of showmount where it is not done this way in the showmount of mentioned package version), or suppling a command line parameter. But I would at least ensure that eigther TCP uder UDP is used for the tool and that this is documented
On 05/08/2013 09:16 PM, EliasOhm wrote:
> I could even post a patch that propably works for both versions, SLES
> and OpenSUSE.
there are no openSUSE 11.x series still in normal support…so we
wouldn’t need that patch…
and, i guess if your collegue with the problem would contact the
SUSE folks they might thank him/her in advance to log a bug and
provide the patch…or maybe they know the problem and how to solve
it without a patch…
so, there is nothing here in these forums for you or the troublesome
SLES 11.1
suggest your next contact be with the correct folks.
As an addition to the fact that these are not the SLES/SLED forums (as allready explained), it is not ver usefull to hang your problem/qusetion at the end of a two years old and long forgotten thread. Almost nobody will be subscribed to it and thus almost nobody will see that you try to enliven it again.
These forums work by advertising your problem. Thus do not hide it at the end of something old, but create a new one in the most fitting (sub) forum with a title that will hopefully atrackt people with knowledge to it.
But this is all theory as you have a SLES/SLED problem.
This thread is CLOSED before it will become a real mess.
