Skip to contentSkip to conversation

AutoFS5

DISCUSSION Jim Summers 3 participants 10 messages Aug 14, 2007 – Aug 15, 2007 Kernel automounter

lore
Share & tools
lore
#0 lore raw

Hello All,

Just hit a hiccup with autofs5. Everything was going along fine and then had a user call and say they could not get their home on a machine. I did some quick checking and some mounts were working and others not. below is a bit of the debug log and some other info. the user, gerardo is the one who called me. entries in the debug log for him are the same as the tmac user. in that it get his ldap info tries to mount but then says it is already mounted and then fails????

client machine is FC6 hand compiled autofs version:
autofs-5.0.1-20
autofs-debuginfo-5.0.1-20
with a patch from Ian.

Yum still isn't pulling the patched one in for me.

==============================

[root@saturn ~]# ls /home
amarack granville

[root@saturn ~]# who

amarack  pts/0        2007-08-13 11:27 (pluto.it.ou.edu)
root     pts/5        2007-08-13 14:49 (redear.it.ou.edu)
gerardo  pts/2        2007-08-14 13:26 (dave.it.ou.edu)
root     pts/3        2007-08-14 13:44 (redear.it.ou.edu)

[root@saturn ~]# su - tmac
su: warning: cannot change directory to /home/tmac: No such file or directory

sun_mount: parse(sun): mounting root /home, mountpoint tmac, what topdawg:/raid/ds9/tmac, fstype nfs, options rw,actimeo=60,rsize=32768,wsize=32768,nocto

Aug 14 14:02:10 saturn automount[2743]: mount_mount: mount(nfs): root=/home 
name=tmac what=topdawg:/raid/ds9/tmac, fstype=nfs, 
options=rw,actimeo=60,rsize=32768,wsize=32768,nocto
Aug 14 14:02:10 saturn automount[2743]: mount_mount: mount(nfs): nfs 
options="rw,actimeo=60,rsize=32768,wsize=32768,nocto", nosymlink=0, ro=0
Aug 14 14:02:10 saturn automount[2743]: mount_mount: mount(nfs): calling 
mkdir_path /home/tmac
Aug 14 14:02:10 saturn automount[2743]: mount_mount: mount(nfs): calling mount 
-t nfs -s -o rw,actimeo=60,rsize=32768,wsize=32768,nocto 
topdawg:/raid/ds9/tmac /home/tmac
Aug 14 14:02:10 saturn automount[2743]: >> mount.nfs: /home/tmac is already 
mounted or busy
Aug 14 14:02:10 saturn automount[2743]: mount(nfs): nfs: mount failure 
topdawg:/raid/ds9/tmac on /home/tmac
Aug 14 14:02:10 saturn automount[2743]: send_fail: token = 284
Aug 14 14:02:10 saturn automount[2743]: failed to mount /home/tmac
Aug 14 14:02:10 saturn automount[2743]: handle_packet: type = 3
Aug 14 14:02:10 saturn automount[2743]: handle_packet_missing_indirect: token 
285, name tmac, request pid 21166
Aug 14 14:02:10 saturn automount[2743]: attempting to mount entry /home/tmac
Aug 14 14:02:10 saturn automount[2743]: lookup_mount: lookup(ldap): looking up 
tmac
Aug 14 14:02:10 saturn automount[2743]: send_fail: token = 285
Aug 14 14:02:10 saturn automount[2743]: failed to mount /home/tmac
Aug 14 14:02:10 saturn automount[2743]: handle_packet: type = 3
Aug 14 14:02:10 saturn automount[2743]: handle_packet_missing_indirect: token 
286, name tmac, request pid 21166
Aug 14 14:02:10 saturn automount[2743]: attempting to mount entry /home/tmac
Aug 14 14:02:10 saturn automount[2743]: lookup_mount: lookup(ldap): looking up 
tmac
Aug 14 14:02:10 saturn automount[2743]: send_fail: token = 286
Aug 14 14:02:10 saturn automount[2743]: failed to mount /home/tmac
Aug 14 14:02:10 saturn automount[2743]: handle_packet: type = 3
Aug 14 14:02:10 saturn automount[2743]: handle_packet_missing_indirect: token 
287, name tmac, request pid 21166
Aug 14 14:02:10 saturn automount[2743]: attempting to mount entry /home/tmac
Aug 14 14:02:10 saturn automount[2743]: lookup_mount: lookup(ldap): looking up 
tmac
Aug 14 14:02:10 saturn automount[2743]: send_fail: token = 287
Aug 14 14:02:10 saturn automount[2743]: failed to mount /home/tmac
Aug 14 14:02:10 saturn automount[2743]: handle_packet: type = 3

===========

ideas / suggestions??

tia

Signature · 3 lines
Jim Summers
School of Computer Science-University of Oklahoma
-------------------------------------------------
#1 lore raw
Quoted 10 lines · Jim Summers

Hello All,

Just hit a hiccup with autofs5. Everything was going along fine and then had a user call and say they could not get their home on a machine. I did some quick checking and some mounts were working and others not. below is a bit of the debug log and some other info. the user, gerardo is the one who called me. entries in the debug log for him are the same as the tmac user. in that it get his ldap info tries to mount but then says it is already mounted and then fails????

Quoted 3 lines

client machine is FC6 hand compiled autofs version:
autofs-5.0.1-20
autofs-debuginfo-5.0.1-20

the rpms were actually built on a FC7 machine and then installed here.

thanks

Quoted 69 lines

with a patch from Ian.

Yum still isn't pulling the patched one in for me.

==============================

[root@saturn ~]# ls /home
amarack granville

[root@saturn ~]# who

amarack  pts/0        2007-08-13 11:27 (pluto.it.ou.edu)
root     pts/5        2007-08-13 14:49 (redear.it.ou.edu)
gerardo  pts/2        2007-08-14 13:26 (dave.it.ou.edu)
root     pts/3        2007-08-14 13:44 (redear.it.ou.edu)

[root@saturn ~]# su - tmac
su: warning: cannot change directory to /home/tmac: No such file or directory

sun_mount: parse(sun): mounting root /home, mountpoint tmac, what topdawg:/raid/ds9/tmac, fstype nfs, options rw,actimeo=60,rsize=32768,wsize=32768,nocto

Aug 14 14:02:10 saturn automount[2743]: mount_mount: mount(nfs): root=/home 
name=tmac what=topdawg:/raid/ds9/tmac, fstype=nfs, 
options=rw,actimeo=60,rsize=32768,wsize=32768,nocto
Aug 14 14:02:10 saturn automount[2743]: mount_mount: mount(nfs): nfs 
options="rw,actimeo=60,rsize=32768,wsize=32768,nocto", nosymlink=0, ro=0
Aug 14 14:02:10 saturn automount[2743]: mount_mount: mount(nfs): calling 
mkdir_path /home/tmac
Aug 14 14:02:10 saturn automount[2743]: mount_mount: mount(nfs): calling mount 
-t nfs -s -o rw,actimeo=60,rsize=32768,wsize=32768,nocto 
topdawg:/raid/ds9/tmac /home/tmac
Aug 14 14:02:10 saturn automount[2743]: >> mount.nfs: /home/tmac is already 
mounted or busy
Aug 14 14:02:10 saturn automount[2743]: mount(nfs): nfs: mount failure 
topdawg:/raid/ds9/tmac on /home/tmac
Aug 14 14:02:10 saturn automount[2743]: send_fail: token = 284
Aug 14 14:02:10 saturn automount[2743]: failed to mount /home/tmac
Aug 14 14:02:10 saturn automount[2743]: handle_packet: type = 3
Aug 14 14:02:10 saturn automount[2743]: handle_packet_missing_indirect: token 
285, name tmac, request pid 21166
Aug 14 14:02:10 saturn automount[2743]: attempting to mount entry /home/tmac
Aug 14 14:02:10 saturn automount[2743]: lookup_mount: lookup(ldap): looking up 
tmac
Aug 14 14:02:10 saturn automount[2743]: send_fail: token = 285
Aug 14 14:02:10 saturn automount[2743]: failed to mount /home/tmac
Aug 14 14:02:10 saturn automount[2743]: handle_packet: type = 3
Aug 14 14:02:10 saturn automount[2743]: handle_packet_missing_indirect: token 
286, name tmac, request pid 21166
Aug 14 14:02:10 saturn automount[2743]: attempting to mount entry /home/tmac
Aug 14 14:02:10 saturn automount[2743]: lookup_mount: lookup(ldap): looking up 
tmac
Aug 14 14:02:10 saturn automount[2743]: send_fail: token = 286
Aug 14 14:02:10 saturn automount[2743]: failed to mount /home/tmac
Aug 14 14:02:10 saturn automount[2743]: handle_packet: type = 3
Aug 14 14:02:10 saturn automount[2743]: handle_packet_missing_indirect: token 
287, name tmac, request pid 21166
Aug 14 14:02:10 saturn automount[2743]: attempting to mount entry /home/tmac
Aug 14 14:02:10 saturn automount[2743]: lookup_mount: lookup(ldap): looking up 
tmac
Aug 14 14:02:10 saturn automount[2743]: send_fail: token = 287
Aug 14 14:02:10 saturn automount[2743]: failed to mount /home/tmac
Aug 14 14:02:10 saturn automount[2743]: handle_packet: type = 3

===========

ideas / suggestions??

tia

Signature · 3 lines
Jim Summers
School of Computer Science-University of Oklahoma
-------------------------------------------------
#2 lore raw
In reply to Jim Summers
Quoted 10 lines · Jim Summers

Hello All,

Just hit a hiccup with autofs5. Everything was going along fine and then had a user call and say they could not get their home on a machine. I did some quick checking and some mounts were working and others not. below is a bit of the debug log and some other info. the user, gerardo is the one who called me. entries in the debug log for him are the same as the tmac user. in that it get his ldap info tries to mount but then says it is already mounted and then fails????

Quoted 4 lines

client machine is FC6 hand compiled autofs version:
autofs-5.0.1-20
autofs-debuginfo-5.0.1-20
with a patch from Ian.

What is the kernel version on the client?

-Jeff

#3 lore raw
Quoted 12 lines · Jeff Moyer
Quoted 10 lines · Jim Summers

Hello All,

Just hit a hiccup with autofs5. Everything was going along fine and then had a user call and say they could not get their home on a machine. I did some quick checking and some mounts were working and others not. below is a bit of the debug log and some other info. the user, gerardo is the one who called me. entries in the debug log for him are the same as the tmac user. in that it get his ldap info tries to mount but then says it is already mounted and then fails????

Quoted 6 lines
Quoted 4 lines

client machine is FC6 hand compiled autofs version:
autofs-5.0.1-20
autofs-debuginfo-5.0.1-20
with a patch from Ian.

What is the kernel version on the client?

I'm guessing that you have 2.6.22.1-13.fc6 or later installed. These kernels include the "nosharecache" patch for NFS. This patch can cause some mounts to fail with -EBUSY. You can revert the nosharecache patch to get the old behaviour back.

Could you let us know what nfs mount options are used for your maps, though, as this should only bite you if you try to mount from the same server file system with different sets of mount options. In other words, your first mount from the server should succeed, successive mount attempts of the same file system with differring options will fail.

-Jeff

#4 lore raw
Quoted 14 lines · Jeff Moyer
Quoted 12 lines · Jeff Moyer
Quoted 10 lines · Jim Summers

Hello All,

Just hit a hiccup with autofs5. Everything was going along fine and then had a user call and say they could not get their home on a machine. I did some quick checking and some mounts were working and others not. below is a bit of the debug log and some other info. the user, gerardo is the one who called me. entries in the debug log for him are the same as the tmac user. in that it get his ldap info tries to mount but then says it is already mounted and then fails????

Quoted 5 lines
Quoted 5 lines
Quoted 4 lines

client machine is FC6 hand compiled autofs version:
autofs-5.0.1-20
autofs-debuginfo-5.0.1-20
with a patch from Ian.

What is the kernel version on the client?

2.6.22.1-32.fc6

Quoted 5 lines

I'm guessing that you have 2.6.22.1-13.fc6 or later installed. These kernels include the "nosharecache" patch for NFS. This patch can cause some mounts to fail with -EBUSY. You can revert the nosharecache patch to get the old behaviour back.

is this accomplished with boot option similar to 'noacpi' or something like that?

Quoted 7 lines

Could you let us know what nfs mount options are used for your maps, though, as this should only bite you if you try to mount from the same server file system with different sets of mount options. In other words, your first mount from the server should succeed, successive mount attempts of the same file system with differring options will fail.

options:

amarak:
-rw,rsize=4096,wsize=4096
here is tmac:
-rw,actimeo=60,rsize=32768,wsize=32768,nocto
here is gerardo:
-rw,rsize=8192,wsize=8192
here is granville:
-rw,rsize=4096,wsize=4096
jsummers:
-rw,actimeo=60,rsize=32768,wsize=32768,nocto

some results:
[root@saturn ~]# su - jsummers
[jsummers@saturn ~]$ exit
logout
[root@saturn ~]# su - tmac
su: warning: cannot change directory to /home/tmac: No such file or directory -bash-3.1$ exit
logout
[root@saturn ~]# su - granville
saturn@~> logout
[root@saturn ~]# su - amarack
saturn@~> logout
[root@saturn ~]# su - gerardo
su: warning: cannot change directory to /home/gerardo: No such file or directory -bash-3.1$ logout

weird, huh?

thanks again

Quoted 7 lines

-Jeff

_______________________________________________
autofs mailing list
[email protected]
http://linux.kernel.org/mailman/listinfo/autofs

#5 lore raw
Quoted 25 lines · Jim Summers
Quoted 15 lines · Jeff Moyer
Quoted 13 lines · Jeff Moyer
Quoted 11 lines · Jim Summers

Hello All,

Just hit a hiccup with autofs5. Everything was going along fine and then had a user call and say they could not get their home on a machine. I did some quick checking and some mounts were working and others not. below is a bit of the debug log and some other info. the user, gerardo is the one who called me. entries in the debug log for him are the same as the tmac user. in that it get his ldap info tries to mount but then says it is already mounted and then fails????

Quoted 5 lines
Quoted 5 lines
Quoted 4 lines

client machine is FC6 hand compiled autofs version:
autofs-5.0.1-20
autofs-debuginfo-5.0.1-20
with a patch from Ian.

What is the kernel version on the client?

2.6.22.1-32.fc6

Quoted 1 line
Quoted 6 lines
Quoted 4 lines

I'm guessing that you have 2.6.22.1-13.fc6 or later installed. These kernels include the "nosharecache" patch for NFS. This patch can cause some mounts to fail with -EBUSY. You can revert the nosharecache patch to get the old behaviour back.

is this accomplished with boot option similar to 'noacpi' or something like that?

Oh, yeah, if you update nfs-utils, you should get access to the nosharecache mount option. You can probably specify this globally in your /etc/sysconfig/autofs file. OPTIONS="-nosharecache", I think.

Quoted 9 lines
Quoted 6 lines

Could you let us know what nfs mount options are used for your maps, though, as this should only bite you if you try to mount from the same server file system with different sets of mount options. In other words, your first mount from the server should succeed, successive mount attempts of the same file system with differring options will fail.

options:

Quoted 10 lines

amarak:
-rw,rsize=4096,wsize=4096
here is tmac:
-rw,actimeo=60,rsize=32768,wsize=32768,nocto
here is gerardo:
-rw,rsize=8192,wsize=8192
here is granville:
-rw,rsize=4096,wsize=4096
jsummers:
-rw,actimeo=60,rsize=32768,wsize=32768,nocto

So, these days superblocks are shared when mounting the same remote file system. As such, when you mount the first export it defines the mount options that will be used for subsequent mounts that come from the same file system.

This is a change in behaviour from what we've had in the past, and there is still ongoing discussion as to how we should resolve the issue.

Quoted 8 lines

some results:
[root@saturn ~]# su - jsummers
[jsummers@saturn ~]$ exit
logout
[root@saturn ~]# su - tmac
su: warning: cannot change directory to /home/tmac: No such file or directory -bash-3.1$ exit
logout

Interesting. I thought that this one would have worked. If you can provide a debug log, we can probably get to the bottom of this. It would also be interesting to know what the contents of /proc/mounts was before each of these commands was run.

For information on collecting a debug log, you can consult my people page:
http://people.redhat.com/jmoyer/

Quoted 9 lines

[root@saturn ~]# su - granville
saturn@~> logout
[root@saturn ~]# su - amarack
saturn@~> logout
[root@saturn ~]# su - gerardo
su: warning: cannot change directory to /home/gerardo: No such file or directory -bash-3.1$ logout

weird, huh?

Definitely. Try the nosharecache mount option. If that resolves the problem, I'm not quite as interested in looking at the debug log.

Thanks!

Jeff

maintainer
#6 lore raw
In reply to Jim Summers
Quoted 24 lines · Jim Summers
Quoted 14 lines · Jeff Moyer
Quoted 12 lines · Jeff Moyer
Quoted 10 lines · Jim Summers

Hello All,

Just hit a hiccup with autofs5. Everything was going along fine and then had a user call and say they could not get their home on a machine. I did some quick checking and some mounts were working and others not. below is a bit of the debug log and some other info. the user, gerardo is the one who called me. entries in the debug log for him are the same as the tmac user. in that it get his ldap info tries to mount but then says it is already mounted and then fails????

Quoted 5 lines
Quoted 5 lines
Quoted 4 lines

client machine is FC6 hand compiled autofs version:
autofs-5.0.1-20
autofs-debuginfo-5.0.1-20
with a patch from Ian.

What is the kernel version on the client?

2.6.22.1-32.fc6

Quoted 1 line
Quoted 6 lines
Quoted 4 lines

I'm guessing that you have 2.6.22.1-13.fc6 or later installed. These kernels include the "nosharecache" patch for NFS. This patch can cause some mounts to fail with -EBUSY. You can revert the nosharecache patch to get the old behaviour back.

is this accomplished with boot option similar to 'noacpi' or something like that?

It's a mount option.
Check your nfs-utils. I believe that revision 14 and above will understand the "nosharecache" option.

Add this as a global option by changing:

#OPTIONS=""

to

OPTIONS="-O nosharecache"

in /etc/sysconfig/autofs, near the bottom and let us know if this helps.

Ian

maintainer
#7 lore raw
Quoted 11 lines · Ian Kent
Quoted 7 lines
Quoted 5 lines

I'm guessing that you have 2.6.22.1-13.fc6 or later installed. These kernels include the "nosharecache" patch for NFS. This patch can cause some mounts to fail with -EBUSY. You can revert the nosharecache patch to get the old behaviour back.

is this accomplished with boot option similar to 'noacpi' or something like that?

It's a mount option.
Check your nfs-utils. I believe that revision 14 and above will understand the "nosharecache" option.

That's nfs-utils-1.0.10-14 for FC6.

If you happen to be using F7 or Rawhide then it's nfs-utils-1.1.0-3, but it doesn't look like the F7 package has got to the mirrors yet. The patch for the option was inadvertently missed from F7 and Rawhide when the patch went into the kernel.

Ian

#8 lore raw
In reply to Ian Kent
Quoted 35 lines · Ian Kent
Quoted 19 lines · Jim Summers
Quoted 18 lines · Jeff Moyer
Quoted 16 lines · Jeff Moyer
Quoted 14 lines · Jim Summers

Hello All,

Just hit a hiccup with autofs5. Everything was going along fine and then had a user call and say they could not get their home on a machine. I did some quick checking and some mounts were working and others not. below is a bit of the debug log and some other info. the user, gerardo is the one who called me. entries in the debug log for him are the same as the tmac user. in that it get his ldap info tries to mount but then says it is already mounted and then fails????

client machine is FC6 hand compiled autofs version:
autofs-5.0.1-20
autofs-debuginfo-5.0.1-20
with a patch from Ian.

Quoted 8 lines
Quoted 1 line
Quoted 1 line

What is the kernel version on the client?

2.6.22.1-32.fc6

Quoted 4 lines

I'm guessing that you have 2.6.22.1-13.fc6 or later installed. These kernels include the "nosharecache" patch for NFS. This patch can cause some mounts to fail with -EBUSY. You can revert the nosharecache patch to get the old behaviour back.

is this accomplished with boot option similar to 'noacpi' or something like that?

It's a mount option.
Check your nfs-utils. I believe that revision 14 and above will understand the "nosharecache" option.

Add this as a global option by changing:

Quoted 8 lines

#OPTIONS=""

to

OPTIONS="-O nosharecache"

in /etc/sysconfig/autofs, near the bottom and let us know if this helps.

seems to have resolved the problem. i added the nosharecache option and restarted the daemon. then was able to su to the affected and unaffected accounts and get the mounts.

the really cool thing was, i was able to do, "/etc/init.d/autofs restart" while a couple of users were logged in and it actually restarted. I was not able to do that with autofs4, it would pretty much hang.

i will monitor it for a few days to verify all is well.

many thanks

Quoted 2 lines

Ian

Signature · 3 lines
Jim Summers
School of Computer Science-University of Oklahoma
-------------------------------------------------
maintainer
#9 lore raw
Quoted 45 lines · Jim Summers
Quoted 35 lines · Ian Kent
Quoted 19 lines · Jim Summers
Quoted 18 lines · Jeff Moyer
Quoted 16 lines · Jeff Moyer
Quoted 14 lines · Jim Summers

Hello All,

Just hit a hiccup with autofs5. Everything was going along fine and then had a user call and say they could not get their home on a machine. I did some quick checking and some mounts were working and others not. below is a bit of the debug log and some other info. the user, gerardo is the one who called me. entries in the debug log for him are the same as the tmac user. in that it get his ldap info tries to mount but then says it is already mounted and then fails????

client machine is FC6 hand compiled autofs version:
autofs-5.0.1-20
autofs-debuginfo-5.0.1-20
with a patch from Ian.

Quoted 8 lines
Quoted 1 line
Quoted 1 line

What is the kernel version on the client?

2.6.22.1-32.fc6

Quoted 4 lines

I'm guessing that you have 2.6.22.1-13.fc6 or later installed. These kernels include the "nosharecache" patch for NFS. This patch can cause some mounts to fail with -EBUSY. You can revert the nosharecache patch to get the old behaviour back.

is this accomplished with boot option similar to 'noacpi' or something like that?

It's a mount option.
Check your nfs-utils. I believe that revision 14 and above will understand the "nosharecache" option.

Add this as a global option by changing:

Quoted 8 lines

#OPTIONS=""

to

OPTIONS="-O nosharecache"

in /etc/sysconfig/autofs, near the bottom and let us know if this helps.

Quoted 7 lines

seems to have resolved the problem. i added the nosharecache option and restarted the daemon. then was able to su to the affected and unaffected accounts and get the mounts.

the really cool thing was, i was able to do, "/etc/init.d/autofs restart" while a couple of users were logged in and it actually restarted. I was not able to do that with autofs4, it would pretty much hang.

Yes, people are starting to notice that.

Ian