autofs from non-Fedora to Fedora 16

Kevin the Drummer · External Since: Nov 19, 2009 Posts: 29

I have a Mandriva 2010.2 system which can NFS mount all of my
other Mandriva systems using autofs to do that. From a command
line all I need do is, for instance

ls /net/machine1/home/

to list the contents of /home on the host machine1. On my Fedora 16
system I can run

ls /net/f16machine/home

and get a listing of /home on the F16 system itself. I believe
that NFS is working because I can run

sudo mount -t nfs f16machine:/home /mnt/F16mount

on a Mandriva machine and then list the contents of /mnt/F16mount
which will show the contents of /home on f16machine.

I can get autofs on the f16machine to negotiate the NFS mount to
my Mandriva machines just fine.

But, when I try to get autofs to negotiate the NFS mount from a
Mandriva machine to the f16machine, then

ls /net/f16machine/home

just times out, and I get a "No such file or directory" error
message on command line. At that time I also get

Apr 11 23:41:58 e129939 automount[19018]: \
lookup_mount: exports lookup failed for f16machine
Apr 11 23:41:58 e129939 automount[19018]: \
update_negative_cache: key "f16machine" not found in map.

in /var/log/messages on the host where I ran the 'ls' command.

Any ideas where/what I should troubleshoot next?

Thanks....

--
PLEASE post a SUMMARY of the answer(s) to your question(s)!
Unless otherwise noted, the statements herein reflect my personal
opinions and not those of any organization with which I may be affiliated.

J G Miller · External Since: Dec 08, 2004 Posts: 360

On Thursday, April 12th, 2012, at 00:38:04h -0700, Kevin the Drummer wrote:

> I have a Mandriva 2010.2 system which can NFS mount all of my
> other Mandriva systems using autofs to do that. From a command
> line all I need do is, for instance

nfs v3 or nfs v4?

In order for nfs v4 to work properly with id mapping you need
different daemons running than for nfs v4, and sometime need
to add an entry to the fstab to get a virtual filesystem
mounted for nfs v4.

So without you indicating which version you are trying to
use and which is available from each machine ...

Kevin the Drummer · External Since: Nov 19, 2009 Posts: 29

J G Miller wrote:
> On Thursday, April 12th, 2012, at 00:38:04h -0700, Kevin the Drummer wrote:
>
> > I have a Mandriva 2010.2 system which can NFS mount all of
> > my other Mandriva systems using autofs to do that. From a
> > command line all I need do is, for instance
>
> nfs v3 or nfs v4?

I suspect that my Fedora system is running NFS v4. I suspect
that my Mandriva systems are running NFS v3. How can I tell for
sure?

> In order for nfs v4 to work properly with id mapping you need
> different daemons running than for nfs v4, and sometime need
> to add an entry to the fstab to get a virtual filesystem
> mounted for nfs v4.

Interesting. I think what you're saying is that autofs makes use
of id mapping, whereas using the 'mount' command with an explicit
host name to be mounted doesn't. That could explain why 'mount'
works for me, while autofs doesn't.

Where should I look for documentation specific to the NFS version
issue surrounding v4 and id mapping?

Is there some compelling reason why I should run v4 instead of
v3? If not, then is there a way to force Fedora 16 to run v3?
I'm presuming here that if I could get everything to run v3, that
this would be easier.

Thanks!

--
PLEASE post a SUMMARY of the answer(s) to your question(s)!
Unless otherwise noted, the statements herein reflect my personal
opinions and not those of any organization with which I may be affiliated.

J G Miller · External Since: Dec 08, 2004 Posts: 360

On Thursday, April 12th, 2012, at 08:25:02h -0700, Kevin the Drummer wrote:

> I suspect that my Fedora system is running NFS v4. I suspect
> that my Mandriva systems are running NFS v3. How can I tell for
> sure?

Possibly by the names or contents of the packages installed.

> Where should I look for documentation specific to the NFS version
> issue surrounding v4 and id mapping?
>
> Is there some compelling reason why I should run v4 instead of
> v3?

Because v4 is actually nicer than v3 if you want to better organise
your exports and because v4 also supports tcp.

For v3 you need rpc.statd running, for v4 you need rpc.idmapd running instead,
as well as rpc_pipefs mounted on /var/lib/nfs/rpc_pipefs with type rpc_pipefs

The format of the exports file for v4 is different and follows a hierarchy

You create a top mount point and export that, eg

/mnt/export *.yourdomain(fsid=0,insecure,no_root_squash,no_subtree_check,ro,sync)

and then export directories under that, eg

/mnt/export/usr/local *.yourdomain(insecure,nohide,no_root_squash,no_subtree_check,ro,sync)

Note that these are for read only exports. Obviously you can create more hierarchies
with fsid=1, fsid=2 etc.

And if you need the security, nfs v4 has support for kerberos.

I have nfsv4 exports set up and use autofs with a custom script inspired
from the autofs auto.net script to do nfs v4 tcp mounting under
/mnt/net/hostname/ with the /mnt/export of the exported share chopped off.

#! /bin/sh

#*****************************************************************************#
#|
#| file : /etc/autofs/auto.export
#|
#*---------------------------------------------------------------------------*#

OPTIONS="-fstype=nfs4,port=2049,proto=tcp"

SHOWMOUNT_FLAGS="--no-headers"

#*---------------------------------------------------------------------------*#

nfs_server="${1}"

#*---------------------------------------------------------------------------*#

showmount=""

for dir in /bin /sbin /usr/bin /usr/sbin
do

for executable in showmount kshowmount
do

file="${dir}/${executable}"

if [ -x "${file}" ]
then
showmount="${file}"
break
fi

done

done

test -x "${showmount}" || exit 1

showmount="${showmount} ${SHOWMOUNT_FLAGS} -e ${nfs_server}"

#*---------------------------------------------------------------------------*#

${showmount} | \
grep '^/mnt/export' | \
grep -v '^/mnt/export[ ]' | \
LC_ALL=C cut -d' ' -f1 | \
LC_ALL=C sort -u | \
awk -v nfs_server="${nfs_server}" -v OPTIONS="${OPTIONS}" -- '
BEGIN {
ORS=""
first=1
}
{
if (first)
{
first=0
print OPTIONS
}
print " \\\n\t" $1, nfs_server ":" $1
}
END {
if (!first)
{
print "\n";
}
else
{
exit 1
}
} ' | \
sed -e 's|/mnt/export/|/|g' -e 's|#|\\#|g'

#*---------------------------------------------------------------------------*#

exit 0

#*****************************************************************************#

There are plenty of tutorials on how to setup nfs v4

<http://www.vanemery.COM/Linux/NFSv4/NFSv4-no-rpcsec.html>

<http://www.novell.COM/coolsolutions/feature/17581.html>

<http://metavo.metacentrum.CZ/en/docs/storage/nfsv4/index.html>

<http://forums.gentoo.ORG/viewtopic-t-748472.html>

Kevin the Drummer · External Since: Nov 19, 2009 Posts: 29

J G Miller wrote:
> On Thursday, April 12th, 2012, at 08:25:02h -0700, Kevin the Drummer wrote:
>
> > I suspect that my Fedora system is running NFS v4. I suspect
> > that my Mandriva systems are running NFS v3. How can I tell for
> > sure?
>
> Possibly by the names or contents of the packages installed.

That was one thing I checked. Names don't infer much to me.

Mandriva 2010.2:
nfs-utils-1.2.2-5.1mdv2010.2
nfs-utils-clients-1.2.2-5.1mdv2010.2

Fedora 16:
nfs-utils-1.2.5-5.fc16.x86_64

Maybe having the nfs-utils files have the same major version
number should be a clue to me that either system could run v4?

> > Where should I look for documentation specific to the NFS version
> > issue surrounding v4 and id mapping?

I made some progress on the idmap issue. On the Mandriva machine
I needed to specify "Domain = myactualdomain.com". Fedora seems
to default OK with that commented out in the idmap.conf file.
Now when I 'sudo mount -t nfs f16machine:/home f16mountdir' and
'ls -l f16mountdir' I get the right user and group names, given
that the UIDs and GIDs match across all my systems.

> > Is there some compelling reason why I should run v4 instead of
> > v3?
>
> Because v4 is actually nicer than v3 if you want to better organise
> your exports and because v4 also supports tcp.

OK. I agree that the exports are a bit of sloppy bookkeeping.
If my network were bigger then I'd have to write some sort of
script to create /etc/exports from some rules, maybe.

> For v3 you need rpc.statd running,

On the Mandriva machine:

ps ax | grep rpc.statd
22378 ? Ss 0:00 rpc.statd --port 4001

> for v4 you need rpc.idmapd running instead,

On the Fedora machine:

systemctl | grep idmap
nfs-idmap.service loaded active running NFSv4 ID-name mapping daemon

> as well as rpc_pipefs mounted on /var/lib/nfs/rpc_pipefs with
> type rpc_pipefs

mount | grep pipe
sunrpc on /var/lib/nfs/rpc_pipefs type rpc_pipefs (rw,relatime)

I think that confirms that I'm running v3 on Mandriva and v4 on Fedora.

> The format of the exports file for v4 is different and follows
> a hierarchy

That might be a source of a problem, given that I treated the
/etc/exports file on the Fedora machine no differently than I do
on my v3 machines.

> You create a top mount point and export that, eg
>
> /mnt/export \
> *.yourdomain(fsid=0,insecure,no_root_squash,no_subtree_check,ro,sync)
>
> and then export directories under that, eg
>
> /mnt/export/usr/local \
> *.yourdomain(insecure,nohide,no_root_squash,no_subtree_check,ro,sync)
>
> Note that these are for read only exports. Obviously you can
> create more hierarchies with fsid=1, fsid=2 etc.

Very interesting. I'll have to dig through the doc's to better
grok that.

> And if you need the security, nfs v4 has support for kerberos.

I'm not too worried about NFS security. The network in my house
is all hardwired. When I go wireless with my laptop, then I use
SSH to connect, and the wireless router only allows connections
with the MAC addresses I've added to its table. It's probably
good for me to start thinking ahead on this though.

> I have nfsv4 exports set up and use autofs with a custom
> script inspired from the autofs auto.net script to do nfs v4
> tcp mounting under /mnt/net/hostname/ with the /mnt/export of
> the exported share chopped off.
>
>
> #! /bin/sh
>
> #************************************************************************#
> #| file : /etc/autofs/auto.export
> #*----------------------------------------------------------------------*#
>
> OPTIONS="-fstype=nfs4,port=2049,proto=tcp"
> SHOWMOUNT_FLAGS="--no-headers"
>
> #*----------------------------------------------------------------------*#
> nfs_server="${1}"
> #*----------------------------------------------------------------------*#
>
> showmount=""
>
> for dir in /bin /sbin /usr/bin /usr/sbin
> do
>
> for executable in showmount kshowmount
> do
>
> file="${dir}/${executable}"
>
> if [ -x "${file}" ]
> then
> showmount="${file}"
> break
> fi
>
> done
>
> done
>
> test -x "${showmount}" || exit 1
>
> showmount="${showmount} ${SHOWMOUNT_FLAGS} -e ${nfs_server}"
>
> #*----------------------------------------------------------------------*#
>
> ${showmount} | \
> grep '^/mnt/export' | \
> grep -v '^/mnt/export[ ]' | \
> LC_ALL=C cut -d' ' -f1 | \
> LC_ALL=C sort -u | \
> awk -v nfs_server="${nfs_server}" -v OPTIONS="${OPTIONS}" -- '
> BEGIN {
> ORS=""
> first=1
> }
> {
> if (first)
> {
> first=0
> print OPTIONS
> }
> print " \\\n\t" $1, nfs_server ":" $1
> }
> END {
> if (!first)
> {
> print "\n";
> }
> else
> {
> exit 1
> }
> } ' | \
> sed -e 's|/mnt/export/|/|g' -e 's|#|\\#|g'
>
> #*----------------------------------------------------------------------*#
> exit 0
> #************************************************************************#
>
>
> There are plenty of tutorials on how to setup nfs v4
>
> <http://www.vanemery.COM/Linux/NFSv4/NFSv4-no-rpcsec.html>

That first one looks excellent.

> <http://www.novell.COM/coolsolutions/feature/17581.html>
> <http://metavo.metacentrum.CZ/en/docs/storage/nfsv4/index.html>
> <http://forums.gentoo.ORG/viewtopic-t-748472.html>

You've been VERY helpful. I really appreciate it. THANKS!

When I get this working, I'll "pay it forward" and post the solution.

--
PLEASE post a SUMMARY of the answer(s) to your question(s)!
Unless otherwise noted, the statements herein reflect my personal
opinions and not those of any organization with which I may be affiliated.

Kevin the Drummer · External Since: Nov 19, 2009 Posts: 29

Kevin the Drummer wrote:
> J G Miller wrote:
> > On Thursday, April 12th, 2012, at 08:25:02h -0700, Kevin the Drummer wrote:
> > The format of the exports file for v4 is different and follows
> > a hierarchy

After reading through all the docs you linked to, it looks like v4 *can*
use a different format, but doesn't have to. To make things simple on
myself I opted to not change this part yet.

> > I have nfsv4 exports set up and use autofs with a custom
> > script inspired from the autofs auto.net script to do nfs v4
> > tcp mounting under /mnt/net/hostname/ with the /mnt/export of
> > the exported share chopped off.

I'll spend some more time with your script. But....

At this point all of the NFSv4 stuff seems to work fine, with the
one annoying exception that autofs won't work. Manually mounting
works completely. It's only autofs that's not working.

I turned up the verbosity of my autofs logging to "debug" and I
get this in /var/log/messages when I try to automount the Fedora
16 machine from the Mandriva machine.

Apr 12 18:40:21 mdvhost automount[17954]: \
st_expire: state 1 path /net
Apr 12 18:40:21 mdvhost automount[17954]: \
expire_proc: exp_proc = 3062889328 path /net
Apr 12 18:40:21 mdvhost automount[17954]: \
expire_cleanup: got thid 3062889328 path /net stat 0
Apr 12 18:40:21 mdvhost automount[17954]: \
expire_cleanup: sigchld: exp 3062889328 finished, switching from 2 to 1
Apr 12 18:40:21 mdvhost automount[17954]: \
st_ready: st_ready(): state = 2 path /net
Apr 12 18:40:34 mdvhost automount[17954]: \
handle_packet: type = 3
Apr 12 18:40:34 mdvhost automount[17954]: \
handle_packet_missing_indirect: \
token 1374, name f16machine, request pid 27782
Apr 12 18:40:34 mdvhost automount[17954]: \
attempting to mount entry /net/f16machine
Apr 12 18:40:34 mdvhost automount[17954]: \
lookup_mount: lookup(program): f16machine -> key is 'f16machine'
Apr 12 18:40:34 mdvhost automount[17954]: \
lookup_mount: lookup(program): looking up f16machine
Apr 12 18:40:34 mdvhost automount[17954]: \
>> clnt_create: RPC: Port mapper failure - \
Unable to receive: errno 113 (No route to host)
Apr 12 18:40:34 mdvhost automount[17954]: \
lookup(program): lookup for f16machine failed
Apr 12 18:40:34 mdvhost automount[17954]: \
dev_ioctl_send_fail: token = 1374
Apr 12 18:40:34 mdvhost automount[17954]: \
failed to mount /net/f16machine
Apr 12 18:40:34 mdvhost automount[17954]: \
handle_packet: type = 3
Apr 12 18:40:34 mdvhost automount[17954]: \
handle_packet_missing_indirect: token 1375, \
name f16machine, request pid 27782
Apr 12 18:40:34 mdvhost automount[17954]: \
attempting to mount entry /net/f16machine
Apr 12 18:40:34 mdvhost automount[17954]: \
dev_ioctl_send_fail: token = 1375
Apr 12 18:40:34 mdvhost automount[17954]: \
failed to mount /net/f16machine
Apr 12 18:41:36 mdvhost automount[17954]: st_expire: state 1 path /net
Apr 12 18:41:36 mdvhost automount[17954]: \
expire_proc: exp_proc = 3062889328 path /net
Apr 12 18:41:36 mdvhost automount[17954]: \
expire_cleanup: got thid 3062889328 path /net stat 0
Apr 12 18:41:36 mdvhost automount[17954]: \
expire_cleanup: sigchld: exp 3062889328 finished, switching from 2 to 1
Apr 12 18:41:36 mdvhost automount[17954]: \
st_ready: st_ready(): state = 2 path /net

Given that my Fedora machine was just installed earlier this
week, I don't have too much time into it. I might try the i686
(or is it i386?) version, rather than the x86_64 version that's
installed right now. There's a little controversy in some forums
as to whether the 64 bit version was broken. I think the expert
opinion was that it's fine. It's a long shot, but I'm stumped
and don't know what else to do.

Thanks....

--
PLEASE post a SUMMARY of the answer(s) to your question(s)!
Unless otherwise noted, the statements herein reflect my personal
opinions and not those of any organization with which I may be affiliated.

J G Miller · External Since: Dec 08, 2004 Posts: 360

On Thursday, April 12th, 2012, at 18:47:24h -0700,
Kevin the Drummer complained:

> At this point all of the NFSv4 stuff seems to work fine, with the
> one annoying exception that autofs won't work.

You are talking about autofs not working with NFSv4 mounts.

How do you think that autofs mounts NFS mounts automagically?

The answer is that it uses whatever you tell it to use in
auto.master, eg in my case for NFSv4 mounts with my script

/mnt/net /etc/autofs/auto.export --timeout 300

So what you need to do is to see which script your automagic
NFS mounts are using, probably the supplied auto.net script
and debug that. That is the reason why I ended up writing
my own script to ensure that the mounts were successful under
/mnt/net/{%hostname}/{%path} rather than
/mnt/net/{%hostname}/export/{%path} or whatever.

You need to issue some smbmount commands manually and then
see how the results from those are being processed by the
supplied auto.net script.

Kevin the Drummer · External Since: Nov 19, 2009 Posts: 29

The problem was on my firewall. The solution is to open ports
111 and 20048 for both udp and tcp. Troubleshooting information
below.

Automounting to the Fedora 16 machine doesn't work yet.

mandriva1> showmount --exports f16machine
rpc mount export: RPC: Unable to receive; errno = No route to host

mandriva1> rpcinfo -p f16machine
rpcinfo: can't contact portmapper: RPC: Remote system error - No route to host

Automounting between Mandriva machines has been working for a while.

mandriva1> showmount --exports mandriva2
Export list for mandriva2:
/spare 192.168.1.0/24
/projects 192.168.1.0/24
/nobackup 192.168.1.0/24
/music 192.168.1.0/24
/media 192.168.1.0/24
/home 192.168.1.0/24

mandriva1> rpcinfo -p mandriva2
program vers proto port
100000 4 tcp 111 portmapper
100000 3 tcp 111 portmapper
100000 2 tcp 111 portmapper
100000 4 udp 111 portmapper
100000 3 udp 111 portmapper
100000 2 udp 111 portmapper
100024 1 udp 4001 status
100024 1 tcp 4001 status
100021 1 udp 4002 nlockmgr
100021 3 udp 4002 nlockmgr
100021 4 udp 4002 nlockmgr
100021 1 tcp 4002 nlockmgr
100021 3 tcp 4002 nlockmgr
100021 4 tcp 4002 nlockmgr
100003 2 tcp 2049 nfs
100003 3 tcp 2049 nfs
100003 4 tcp 2049 nfs
100227 2 tcp 2049 nfs_acl
100227 3 tcp 2049 nfs_acl
100003 2 udp 2049 nfs
100003 3 udp 2049 nfs
100003 4 udp 2049 nfs
100227 2 udp 2049 nfs_acl
100227 3 udp 2049 nfs_acl
100005 1 udp 33675 mountd
100005 1 tcp 59528 mountd
100005 2 udp 33675 mountd
100005 2 tcp 59528 mountd
100005 3 udp 33675 mountd
100005 3 tcp 59528 mountd

I opened port 111 for tcp and udp on the f16machine firewall. Then I got:

mandriva1> rpcinfo -p f16machine
program vers proto port
100000 4 tcp 111 portmapper
100000 3 tcp 111 portmapper
100000 2 tcp 111 portmapper
100000 4 udp 111 portmapper
100000 3 udp 111 portmapper
100000 2 udp 111 portmapper
100011 1 udp 875 rquotad
100011 2 udp 875 rquotad
100011 1 tcp 875 rquotad
100011 2 tcp 875 rquotad
100024 1 udp 4001 status
100024 1 tcp 4001 status
100003 2 tcp 2049 nfs
100003 3 tcp 2049 nfs
100003 4 tcp 2049 nfs
100227 2 tcp 2049 nfs_acl
100227 3 tcp 2049 nfs_acl
100003 2 udp 2049 nfs
100003 3 udp 2049 nfs
100003 4 udp 2049 nfs
100227 2 udp 2049 nfs_acl
100227 3 udp 2049 nfs_acl
100021 1 udp 4002 nlockmgr
100021 3 udp 4002 nlockmgr
100021 4 udp 4002 nlockmgr
100021 1 tcp 4002 nlockmgr
100021 3 tcp 4002 nlockmgr
100021 4 tcp 4002 nlockmgr
100005 1 udp 20048 mountd
100005 1 tcp 20048 mountd
100005 2 udp 20048 mountd
100005 2 tcp 20048 mountd
100005 3 udp 20048 mountd
100005 3 tcp 20048 mountd

But, still I had no luck.

mandriva1> showmount --exports f16machine
rpc mount export: RPC: Unable to receive; errno = No route to host

I tried opening all of the ports listed in rpcinfo on the
firewall. That worked. Then I backed off to opening only ports
111 and 20048, both for tcp and udp. Now I can see the mount
points.

mandriva1> showmount --exports f16machine
Export list for f16machine:
/projects 192.168.1.0/24
/nobackup 192.168.1.0/24
/music 192.168.1.0/24
/home 192.168.1.0/24

What's better is that I can mount the mount points!

Ahhh, nice. I hope this helps someone else!

Cheers.

--
PLEASE post a SUMMARY of the answer(s) to your question(s)!
Unless otherwise noted, the statements herein reflect my personal
opinions and not those of any organization with which I may be affiliated.

J G Miller · External Since: Dec 08, 2004 Posts: 360

On Saturday, April 14th, 2012, at 22:57:11h -0700, Kevin the Drummer wrote:

> The problem was on my firewall.

But which firewall? The one on Fedora machine, the one
on the Mageia machine, or the machine doing the routing
if not a consumer home router?

> mandriva1> showmount --exports f16machine
> rpc mount export: RPC: Unable to receive; errno = No route to host

That was why I suggested you first use showmount on the remote
machine to ensure that you could actually see the exports.

Your next task should be to get it all working with just nfs v4 Wink

Kevin the Drummer · External Since: Nov 19, 2009 Posts: 29

J G Miller wrote:
> On Saturday, April 14th, 2012, at 22:57:11h -0700, Kevin the Drummer wrote:
>
> > The problem was on my firewall.
>
> But which firewall? The one on Fedora machine, the one
> on the Mageia machine, or the machine doing the routing
> if not a consumer home router?

It was the firewall on the Fedora (server) machine. I didn't
have to touch any of the client firewalls. The two machines are
connected via a switch. My router is on the other side of the
switch from the machines.

> > mandriva1> showmount --exports f16machine
> > rpc mount export: RPC: Unable to receive; errno = No route to host
>
> That was why I suggested you first use showmount on the remote
> machine to ensure that you could actually see the exports.
>
> Your next task should be to get it all working with just nfs v4 Wink

I might just try that when I upgrade a 2nd machine to Fedora 16.

Thanks....

--
PLEASE post a SUMMARY of the answer(s) to your question(s)!
Unless otherwise noted, the statements herein reflect my personal
opinions and not those of any organization with which I may be affiliated.