autofs4 with NetApp-Filer and large export list: Stopped working for me
Hi,
I used to use autofs4 with a NetApp-Filer as NFS server (mostly from SLES9 boxes). This does not seem to work any longer. So I digged a little bit deeper:
The NetApp-Filer in question exports nearly 100 volumes:
sappscmn:/root>>> showmount --no-headers -e 10.172.230.134 | wc -l 96
All of them have access restrictions:
sappscmn:/root>>> showmount --no-headers -e 10.172.230.134 | awk '$2==""' sappscmn:/root>>>
sappscmn:/root>>> showmount --no-headers -e 10.172.230.134 | head -5
/vol/nb/sles8/shared_os/var_img/var-0aac118a 10.172.17.138 /vol/nb/sles9_ia32/shared_os/root_aclm 10.172.17.99 /vol/nb/sles9_ia32/shared_os/var_img/var-0aac1163 10.172.17.99
/vol/nb/sles9_ia32/shared_os/root_cmsa_p
10.172.17.122,10.172.17.123
/vol/nb/sles9_ia32/shared_os/var_img/var-0aac1164 10.172.17.100
My IP-address is
sappscmn:/root>>> ip route list dev eth1 10.172.230.0/24 proto kernel scope link src 10.172.230.142 169.254.0.0/16 scope link
There is one volume accessible from the current host:
sappscmn:/root>>> showmount --no-headers -e 10.172.230.134 | grep
10.172.230.0/24
/vol/pcc 172.18.144.122,172.18.11.179,172.18.214.253,172.18.144.64/26,172.18.139.0/26,10.172.139.0/26,172.18.230.0/24,10.172.230.0/24,10.172.17.0/24,10.172.144.128/26,172.24.207.0/24,172.18.139.10,172.18.30.224/27,172.18.102.0/24,172.18.107.0/24,172.18.95.0/24,172.18.228.0/24,10.172.144.192/26,10.172.144.64/26,172.18.131.0/24,10.125.1.35,172.18.28.0/24,10.172.143.0/27,172.18.170.0/24,172.18.117.133,172.18.95.180,172.18.229.0/24,172.18.30.192/27,172.18.158.0/24,172.18.96.128/25 sappscmn:/root>>>
No other volume may be mounted.
Nevertheless automount tries to mount all volumes.
Running /etc/auto.net gives me:
sappscmn:/root>>> /etc/auto.net 10.172.230.134
-fstype=nfs,hard,intr,nodev,nosuid \
/vol/acid 10.172.230.134:/vol/acid \
/vol/casm_log/casm_log 10.172.230.134:/vol/casm_log/casm_log \
/vol/casm_log 10.172.230.134:/vol/casm_log \
/vol/casm 10.172.230.134:/vol/casm \
...
/vol/pcc 10.172.230.134:/vol/pcc \...
sappscmn:/root>>> /etc/init.d/autofs start
Starting service automounter done
sappscmn:/root>>> cd /net/10.172.230.134/vol/pcc
bash: cd: /net/10.172.230.134/vol/pcc: No such file or directory sappscmn:/root>>>
/var/log/messages shows quite clearly:
Apr 27 11:25:05 sappscmn automount[7822]: >> mount: 10.172.230.134:/vol/acid failed, reason given by server: Permission denied Apr 27 11:25:05 sappscmn automount[7822]: mount(nfs): nfs: mount failure 10.172.230.134:/vol/acid on /net/10.172.230.134/vol/acid Apr 27 11:25:05 sappscmn automount[7822]: >> mount:
10.172.230.134:/vol/casm_log/casm_log failed, reason given by server:
Permission denied
Up to this point there are no news. This happened always.
Unfortunately the volume I want to access cannot be mounted any longer:
...
Apr 27 11:25:06 sappscmn automount[7822]: >> mount: 10.172.230.134:/vol/nb/sles9_ia32/shared_os/root_cms_p failed, reason given by server: Permission denied Apr 27 11:25:06 sappscmn automount[7822]: mount(nfs): nfs: mount failure 10.172.230.134:/vol/nb/sles9_ia32/shared_os/root_cms_p on /net/10.172.230.134/vol/nb/sles9_ia32/shared_os/root_cms_p Apr 27 11:25:06 sappscmn automount[7822]: >> mount: fs type, bad option, bad superblock ... or too many mounted file systems Apr 27 11:25:06 sappscmn automount[7876]: >> /usr/sbin/showmount: can't get address for 10.172.230.134/vol/pcc Apr 27 11:25:06 sappscmn automount[7876]: lookup(program): lookup for 10.172.230.134/vol/pcc failed Apr 27 11:25:06 sappscmn automount[7876]: failed to mount
/net/10.172.230.134/vol/pcc
If I change /etc/auto.net from
$SHOWMOUNT | LC_ALL=C sort -k 1 | \
to
$SHOWMOUNT | LC_ALL=C sort -k 1 | grep 10.172.230.0/24 | \
everything works well.
But this is a hack.
So my questions:
Is it possible make showmount show only volumes, I could mount succesful and to suppress those, that will fail ?
Are there know problems with large export list (It used to work before, but the NetApp-Filer became bigger and bigger ...)
Any other hints ?
Thanks for your time !
Kind regards
Gerd
==> On Fri, 27 Apr 2007 14:33:24 +0200, Gerd Bavendiek <[email protected]> said:
Gerd> Hi,
Gerd> I used to use autofs4 with a NetApp-Filer as NFS server (mostly from Gerd> SLES9 boxes). This does not seem to work any longer. So I digged a Gerd> little bit deeper:
Thanks for the detailed problem report!
Gerd> The NetApp-Filer in question exports nearly 100 volumes:
[snip]
Gerd> Apr 27 11:25:06 sappscmn automount[7876]: >> /usr/sbin/showmount: can't Gerd> get address for 10.172.230.134/vol/pcc
This is a problem. This means that the kernel is passing along a lookup for 10.172.230.134/vol/pcc, when it should only be asking for 10.172.230.134. What kernel version are you running? Can you update to the latest vendor-supplied kernel and try again? (I'm not sure if SuSE fixed this in their kernels, but I'm hoping they did).
If the latest kernel does not fix your problem, please respond here and we'll point you at the appropriate patches to apply.
Note that autofs version 5 now implements a 'hosts' lookup module which is a bit less kludgy than auto.net. Ian may have something to add to this conversation regarding the access lists.
-Jeff
Ian Kent
maintainerQuoted 98 lines · Gerd Bavendiek
Hi,
I used to use autofs4 with a NetApp-Filer as NFS server (mostly from SLES9 boxes). This does not seem to work any longer. So I digged a little bit deeper:
The NetApp-Filer in question exports nearly 100 volumes:
sappscmn:/root>>> showmount --no-headers -e 10.172.230.134 | wc -l 96
All of them have access restrictions:
sappscmn:/root>>> showmount --no-headers -e 10.172.230.134 | awk '$2==""' sappscmn:/root>>>
sappscmn:/root>>> showmount --no-headers -e 10.172.230.134 | head -5/vol/nb/sles8/shared_os/var_img/var-0aac118a 10.172.17.138 /vol/nb/sles9_ia32/shared_os/root_aclm 10.172.17.99 /vol/nb/sles9_ia32/shared_os/var_img/var-0aac1163 10.172.17.99/vol/nb/sles9_ia32/shared_os/root_cmsa_p
10.172.17.122,10.172.17.123
/vol/nb/sles9_ia32/shared_os/var_img/var-0aac1164 10.172.17.100My IP-address is
sappscmn:/root>>> ip route list dev eth1 10.172.230.0/24 proto kernel scope link src 10.172.230.142 169.254.0.0/16 scope link
There is one volume accessible from the current host:
sappscmn:/root>>> showmount --no-headers -e 10.172.230.134 | grep
10.172.230.0/24
/vol/pcc 172.18.144.122,172.18.11.179,172.18.214.253,172.18.144.64/26,172.18.139.0/26,10.172.139.0/26,172.18.230.0/24,10.172.230.0/24,10.172.17.0/24,10.172.144.128/26,172.24.207.0/24,172.18.139.10,172.18.30.224/27,172.18.102.0/24,172.18.107.0/24,172.18.95.0/24,172.18.228.0/24,10.172.144.192/26,10.172.144.64/26,172.18.131.0/24,10.125.1.35,172.18.28.0/24,10.172.143.0/27,172.18.170.0/24,172.18.117.133,172.18.95.180,172.18.229.0/24,172.18.30.192/27,172.18.158.0/24,172.18.96.128/25 sappscmn:/root>>>No other volume may be mounted.
Nevertheless automount tries to mount all volumes.
Running
/etc/auto.netgives me:sappscmn:/root>>>
/etc/auto.net10.172.230.134-fstype=nfs,hard,intr,nodev,nosuid \ /vol/acid 10.172.230.134:/vol/acid \ /vol/casm_log/casm_log 10.172.230.134:/vol/casm_log/casm_log \ /vol/casm_log 10.172.230.134:/vol/casm_log \ /vol/casm 10.172.230.134:/vol/casm \ ... /vol/pcc 10.172.230.134:/vol/pcc \...
sappscmn:/root>>>
/etc/init.d/autofsstart
Starting service automounter done
sappscmn:/root>>> cd /net/10.172.230.134/vol/pcc
bash: cd: /net/10.172.230.134/vol/pcc: No such file or directory sappscmn:/root>>>
/var/log/messagesshows quite clearly:Apr 27 11:25:05 sappscmn automount[7822]: >> mount: 10.172.230.134:/vol/acid failed, reason given by server: Permission denied Apr 27 11:25:05 sappscmn automount[7822]: mount(nfs): nfs: mount failure 10.172.230.134:/vol/acid on /net/10.172.230.134/vol/acid Apr 27 11:25:05 sappscmn automount[7822]: >> mount:10.172.230.134:/vol/casm_log/casm_log failed, reason given by server:
Permission deniedUp to this point there are no news. This happened always.
Unfortunately the volume I want to access cannot be mounted any longer:
...
Apr 27 11:25:06 sappscmn automount[7822]: >> mount: 10.172.230.134:/vol/nb/sles9_ia32/shared_os/root_cms_p failed, reason given by server: Permission denied Apr 27 11:25:06 sappscmn automount[7822]: mount(nfs): nfs: mount failure 10.172.230.134:/vol/nb/sles9_ia32/shared_os/root_cms_p on /net/10.172.230.134/vol/nb/sles9_ia32/shared_os/root_cms_p Apr 27 11:25:06 sappscmn automount[7822]: >> mount: fs type, bad option, bad superblock ... or too many mounted file systems Apr 27 11:25:06 sappscmn automount[7876]: >> /usr/sbin/showmount: can't get address for 10.172.230.134/vol/pcc Apr 27 11:25:06 sappscmn automount[7876]: lookup(program): lookup for 10.172.230.134/vol/pcc failed Apr 27 11:25:06 sappscmn automount[7876]: failed to mount/net/10.172.230.134/vol/pcc
If I change
/etc/auto.netfrom$SHOWMOUNT | LC_ALL=C sort -k 1 | \
to
$SHOWMOUNT | LC_ALL=C sort -k 1 | grep 10.172.230.0/24 | \
Quoted 8 lines
everything works well.
But this is a hack.
So my questions:
Is it possible make showmount show only volumes, I could mount succesful and to suppress those, that will fail ?
This is a contentious issue.
In version 5 exports for which the client didn't have access were pruned for a long time but since the validation of exports can't work from behind a NAT router I've disabled it in a recent patch.
So you'll still have to put up with those annoying messages.
Quoted 3 lines
Are there know problems with large export list (It used to work before, but the NetApp-Filer became bigger and bigger ...)
Yes, several.
Quoted 2 lines
Any other hints ?
Use version 5.
The latest version is available on kernel.org.
Use all the patches for 5.0.1 as well.
If this works for you then bug your vendor about updating to v5.
Ian