NFS install problem
in [Slackware]
|
Next: tor
From: Mark Johnson on 5 Feb 2005 22:54 Hi, While trying to install slackware 10.0 via NFS, setup appears to go through all the steps (sets up swap, selects root partition, packages, etc), but then fails to install any packages. Furthermore, setup fails to detect that this is an error and prompts me to finish configuring the system. The entire "Full install" takes only ten seconds or so. Afterwards, there are no files on the drive. Only a few directories. This happens whether I choose to install via NFS or mount the slackware NFS drive and choose to install from a premounted volume. The only clue I have is that when I manually mount the NFS drive, it prints the following error: RPC: call_verify: program 100003, version 3 unsupported by server 192.168.1.3 nfs_get_root: getattr error = 5 Now, the server should support NFS version 3 since I've recompiled the kernel to do so. The server is running Slackware 10,0, with the kernel upgraded to 2.6.10, and a few security patches applied. Additionally, despite the error, the client appears to have successfully mounted the drive. I can browse all the directories. I can even run installpkg (to install packages from the NFS) from the command line successfully. It's only setup that fails to install files. Any help? Mark
From: Henrik Carlqvist on 6 Feb 2005 07:41 Mark Johnson <mXrXj001(a)shaw.ca> wrote: > While trying to install slackware 10.0 via NFS, setup appears to go > through all the steps (sets up swap, selects root partition, packages, > etc), but then fails to install any packages. I have no experience from Slackware 10 myself, but have done several Slackware 9.1 nfs installations. > This happens whether I choose to install via NFS or mount the slackware > NFS drive and choose to install from a premounted volume. Are you sure that it is mounted in the right directory where the setup scripts will be looking for packages? It is important to mount the right level on the right direktory. The level to mount is the directory containing the package directories (a, ap and so on). So, lets say you have a server with something like the following in /etc/exports: /nfs-exports all(ro) And the server has the following directory (among others) /nfs-exports/slack10/a Then it is important that you do the following mount: mount myserver:/nfs-exports/slack10 / /var/log/mount If you instead do: mount myserver:/nfs-exports /var/log/mount the installation scripts will not be able to find your packages. > The only clue I have is that when I manually mount the NFS drive, it > prints the following error: > RPC: call_verify: program 100003, version 3 unsupported by server > 192.168.1.3 > nfs_get_root: getattr error = 5 NFS is supposed to be backwards compatible. Even if you only get NFS version 2 things should work. You could verify which version of NFS you get by doing "cat /proc/mounts". If you have a field saying "v3" you have NFS version 3, if it says "v2" it is only NFS version 2. > Now, the server should support NFS version 3 since I've recompiled the > kernel to do so. As NFS version 3 is faster you might want to try to get it working. regards Henrik -- The address in the header is only to prevent spam. My real address is: hc2(at)uthyres.com Examples of addresses which go to spammers: info(a)k-software.biz svar(a)webtelevision.se info(a)webrider.ru root(a)localhost
From: Mark Johnson on 6 Feb 2005 10:48 Henrik Carlqvist wrote: > Mark Johnson <mXrXj001(a)shaw.ca> wrote: [snip] > > Are you sure that it is mounted in the right directory where the setup > scripts will be looking for packages? It is important to mount the right > level on the right direktory. The level to mount is the directory > containing the package directories (a, ap and so on). > Setup actually provides two options you can use for this. One is to let it mount the NFS volume, and the other is to install from a premounted volume. I have tried both. Neither worked. The failure was the same each time. > So, lets say you have a server with something like the following in > /etc/exports: > > /nfs-exports all(ro) > > And the server has the following directory (among others) > > /nfs-exports/slack10/a > > Then it is important that you do the following mount: > > mount myserver:/nfs-exports/slack10 / /var/log/mount > > If you instead do: > > mount myserver:/nfs-exports /var/log/mount > > the installation scripts will not be able to find your packages. I tried both of these. > > >>The only clue I have is that when I manually mount the NFS drive, it >>prints the following error: >>RPC: call_verify: program 100003, version 3 unsupported by server >>192.168.1.3 >>nfs_get_root: getattr error = 5 > > > NFS is supposed to be backwards compatible. Even if you only get NFS > version 2 things should work. You could verify which version of NFS you > get by doing "cat /proc/mounts". If you have a field saying "v3" you have > NFS version 3, if it says "v2" it is only NFS version 2. > This does show v2. The error says that the server does not support v3. However, I compiled v3 support into the kernel. Is there anything else I need to do to support v3? Could it really be that the client (bare.i kernel) does not support v3? > > As NFS version 3 is faster you might want to try to get it working. > Actually, I am much more concerned with getting the install working than with getting v3 of NFS working. Even when I let setup do everything regarding mounting the NFS, I get a "full install" that takes about 5 seconds and installs nothing. This is the real problem, not which version of NFS I am using. The second time, I ran "setup 2>log". It showed the following: /usr/lib/setup/SeTswap: /proc/meminfo: Permission denied SIOCADDRT: File exists chmod: /mnt/tmp: No such file or directory On this run, I didn't let setup run mkswap (the first run had done that), and I didn't format my root partition. Mark
From: Henrik Carlqvist on 6 Feb 2005 11:22 Mark Johnson <mXrXj001(a)shaw.ca> wrote: >> Are you sure that it is mounted in the right directory where the setup >> scripts will be looking for packages? > I tried both of these. Then please double-check where the Slackware install script is looking for the packages. As you are able to manually install packages there seems to be something wrong with the script or where you mounted what. > This does show v2. The error says that the server does not support v3. > However, I compiled v3 support into the kernel. Is there anything > else I need to do to support v3? Could it really be that the client > (bare.i kernel) does not support v3? I have no experience from the 2.6 kernel you are using. However, the 2.4 kernel has two options: CONFIG_NFS_V3 CONFIG_NFSD_V3 With the first option your client will be able to do NFS v3 mounts. With the second option your server will be able to export NFS v3 mounts. It is a good idea to have both enabled. > Actually, I am much more concerned with getting the install working Yes, I can understand that. "Use the source luke!" Find out which installation scripts are running, read them and try to understand exactly what they are trying to do. I still think that they failed bacause you did mount the wrong directory or did mount it at the wrong place, but maybe there is some bug in the Slack 10 nfs installation scripts. If you read the scripts and understand them you will find that bug or find out what you did wrong. regards Henrik -- The address in the header is only to prevent spam. My real address is: hc2(at)uthyres.com Examples of addresses which go to spammers: info(a)k-software.biz svar(a)webtelevision.se info(a)webrider.ru root(a)localhost
From: GrantC on 6 Feb 2005 15:59
Hi there On Sun, 06 Feb 2005 15:48:44 GMT, Mark Johnson <mXrXj001(a)shaw.ca> wrote: >Henrik Carlqvist wrote: >> Mark Johnson <mXrXj001(a)shaw.ca> wrote: >[snip] >> >> Are you sure that it is mounted in the right directory where the setup >> scripts will be looking for packages? It is important to mount the right >> level on the right direktory. The level to mount is the directory >> containing the package directories (a, ap and so on). first gotcha: Just adding what works for me, on the server: peetoo:~$ cat /etc/exports # See exports(5) for a description. # This file contains a list of all directories exported to other computers. # It is used by rpc.nfsd and rpc.mountd. /home/install 192.168.1.0/24(sync,rw,no_root_squash) That 'sync' option needs to be explicit for recent versions of NFS. The source NFS mountpoint is specified in terms of the server path. Any errors at the source mounting stage must be resolved, you specify source directory (from memory) from the server root, for example, my copy of slackware-current sits in /home/install, so I'd try: /home/install/slackware-current/slackware If the source isn't seen, the slack installer will offer to retry, do that a few times until you hit the right incantation that doesn't throw an error at you. second gotcha: If you mount the source yourself, create a separate mountpoint on the target root, this is explained somewhere in the install docs. I create my own /src and /dst for mucking around the target filesystem 'outside' of the slackware installer's target or working area. Hope this helps. Cheers, Grant. |