Skip to contentSkip to conversation

Help with automounting from a LDAP server

DISCUSSION Krishnakumar B 5 participants 8 messages Jan 10, 2005 – Jan 12, 2005 Kernel automounter

lore
Share & tools
lore
#0 lore raw

Hi,

Are there any standard practices to follow for automounting from a LDAP server without using local files, for both the NFS server and the clients? The NFS clients part seems straightforward. However it is not clear how autofs mounts directories residing on the local filesystem on the NFS server.

For example if I have the following setup:

1. I export /export/home from the NFS server, nfs.example.com
2. The NFS server is also configured to use the LDAP server to get automount information, i.e, automount: ldap in /etc/nsswitch.conf
3. auto_home is setup on the LDAP server to spit out something equivalent to:

* -rw,hard,intr,rsize=8192,wsize=8192,nosuid nfs.example.com:/export/home

Will autofs on nfs.example.com recognize the local mount /export/home and use a loopback mount for /home? If so, do people recommend using the same approach (using a direct mount) for /var/spool/mail on nfs.example.com? If not, why not?

Is there any way to disable specify exclude clauses in the mount options to autofs? For example, can I specify some option to autofs so that on a particular machine autofs will not use the automount entry from the directory server.

Thanks,
kitty.

Signature · 2 lines
Krishnakumar B <kitty at dre dot vanderbilt dot edu>
Institute for Software Integrated Systems, Dept. of EECS, Vanderbilt University
maintainer
#1 lore raw
Quoted 7 lines · Krishnakumar B

Hi,

Are there any standard practices to follow for automounting from a LDAP server without using local files, for both the NFS server and the clients? The NFS clients part seems straightforward. However it is not clear how autofs mounts directories residing on the local filesystem on the NFS server.

autofs should perform a bind mount if the mount is on the local machine.

Quoted 8 lines

For example if I have the following setup:

1. I export /export/home from the NFS server, nfs.example.com
2. The NFS server is also configured to use the LDAP server to get automount information, i.e, automount: ldap in /etc/nsswitch.conf
3. auto_home is setup on the LDAP server to spit out something equivalent to:

Quoted 6 lines

* -rw,hard,intr,rsize=8192,wsize=8192,nosuid nfs.example.com:/export/home

Will autofs on nfs.example.com recognize the local mount /export/home and use a loopback mount for /home? If so, do people recommend using the same approach (using a direct mount) for /var/spool/mail on nfs.example.com? If not, why not?

Can't do that at present.
Direct mounts are not fully implemented.

Quoted 5 lines

Is there any way to disable specify exclude clauses in the mount options to autofs? For example, can I specify some option to autofs so that on a particular machine autofs will not use the automount entry from the directory server.

Not that I know of.

Ian

#2 lore raw

Hi,

Thanks for all the followups.

Quoted 2 lines · Ian Kent

On Mon, 10 Jan 2005, Krishnakumar B wrote:

[...]

Quoted 6 lines
Quoted 4 lines

Will autofs on nfs.example.com recognize the local mount /export/home and use a loopback mount for /home? If so, do people recommend using the same approach (using a direct mount) for /var/spool/mail on nfs.example.com? If not, why not?

Can't do that at present. Direct mounts are not fully implemented.

My manpage for autofs(5) claims the following:

UNSUPPORTED
This version of the automounter supports direct maps for FILE, NIS and LDAP maps only and handles SunOS-style replicated filesystems only to the extent that mount(8) does.

So is my manpage wrong, or does "direct maps" mean something other than "direct mounts"? Or are you referring to the following limitation (from README.direct v1.3).

NOTE: Due to current design limitations, direct maps will take over an entire directory hierarchy. What this means is, if your direct map key is /usr/share/bilbo, then /usr will become an automount mount point, mounting over the existing /usr.

I am using the following version of autofs with Linux-2.4.28.

tango % dpkg -S /usr/share/man/man5/autofs.5.gz
autofs: /usr/share/man/man5/autofs.5.gz
tango %

tango % dpkg -p autofs
Package: autofs
Priority: extra
Section: utils
Installed-Size: 428
Maintainer: Steinar H. Gunderson <[email protected]>
Architecture: i386
Version: 4.1.3-8
DDepends: libc6 (>= 2.3.2.ds1-4)
Recommends: nfs-common
ConCflicts: samba (<< 2.0.6-1)
Filename: pool/main/a/autofs/autofs_4.1.3-8_i386.deb
Size: 96350
MD5sum: 42aef00313ebe91c572bcdeb04708914
Description: A kernel-based automounter for Linux
Autofs controls the operation of the automount daemons. The automount daemons automatically mount filesystems when they are used and unmount them after a period of inactivity. This is done based on a set of pre-configured maps.
.
The kernel automounter implements an almost complete SunOS style automounter under Linux. Automounter version 4 (autofs4) has to be enabled when compiling the kernel. Debian packaged kernels have it enabled.

-kitty.

Signature · 2 lines
Krishnakumar B <kitty at dre dot vanderbilt dot edu>
Institute for Software Integrated Systems, Dept. of EECS, Vanderbilt University
maintainer
#3 lore raw
Quoted 8 lines · Krishnakumar B
Quoted 1 line

Can't do that at present. Direct mounts are not fully implemented.

My manpage for autofs(5) claims the following:

UNSUPPORTED
This version of the automounter supports direct maps for FILE, NIS and LDAP maps only and handles SunOS-style replicated filesystems only to the extent that mount(8) does.

Perhaps I need to change this to be more accurate.

Quoted 9 lines

So is my manpage wrong, or does "direct maps" mean something other than "direct mounts"? Or are you referring to the following limitation (from README.direct v1.3).

NOTE: Due to current design limitations, direct maps will take over an entire directory hierarchy. What this means is, if your direct map key is /usr/share/bilbo, then /usr will become an automount mount point, mounting over the existing /usr.

I'm refering to the design limitation mentioned in this note.

This will continue to be the case in 4.1.4 (in fact 4.1.x).

I plan on creating a 4.2.0 development branch at some point fairly soon. This will be the top priority for that development.

Ian

#4 lore raw

sorry about this newbie question, but can you explain what exactly direct mounts/maps is?

thx

Quoted 3 lines · On Tue, 11 Jan 2005 16:39:40 +0800 (WST) Ian Kent <[email protected]> wrote:
Quoted 1 line · Krishnakumar B
Quoted 1 line

Can't do that at present. Direct mounts are not fully

implemented.

Quoted 5 lines
Quoted 5 lines

My manpage for autofs(5) claims the following:

UNSUPPORTED
This version of the automounter supports direct maps for

FILE, NIS

Quoted 1 line
Quoted 1 line

and LDAP maps only and handles SunOS-style replicated

filesystems

Quoted 6 lines
Quoted 1 line

only to the extent that mount(8) does.

Perhaps I need to change this to be more accurate.

Quoted 2 lines

So is my manpage wrong, or does "direct maps" mean something other

than

Quoted 1 line
Quoted 1 line

"direct mounts"? Or are you referring to the following limitation

(from

Quoted 3 lines
Quoted 3 lines

README.direct v1.3).

NOTE: Due to current design limitations, direct maps will take over

an

Quoted 1 line
Quoted 1 line

entire directory hierarchy. What this means is, if your direct map

key is

Quoted 1 line
Quoted 1 line

/usr/share/bilbo, then /usr will become an automount mount point,

mounting

Quoted 7 lines
Quoted 1 line

over the existing /usr.

I'm refering to the design limitation mentioned in this note.

This will continue to be the case in 4.1.4 (in fact 4.1.x).

I plan on creating a 4.2.0 development branch at some point fairly

soon.

Quoted 8 lines

This will be the top priority for that development.

Ian

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

#5 lore raw
Quoted 2 lines · abo

sorry about this newbie question, but can you explain what exactly direct mounts/maps is?

That's a little difficult but ...

An example of an indirect map might be:

in the master map

/home /etc/auto.indirect

and auto.indirect might have

user1 server1:/home/userdir1
user2 server1:/home/userdir2
....

An example of a direct map might be:

/- /etc/auto.direct

and in auto.direct might have

/nfs/apps/geoapp1 server1:/apps/geoapplication
/nfs/apps/geoapp2 server2:/apps/application2
/nfs/data/mydata data1:/data/north/mydata
/global/data/theirdata other:/data/test
...

Direct mount maps always have /- in the master map whereas indirect maps always have their base directory. Direct mount keys are always full paths wheras in an indirect mount the key is always a single directory component.

You can't have nested mounts in a direct map either.

Check out Managing NFS and NIS 2nd edition, O'Reilly, chapter 13.

Quoted 58 lines

thx

Quoted 3 lines · On Tue, 11 Jan 2005 16:39:40 +0800 (WST) Ian Kent <[email protected]> wrote:
Quoted 1 line · Krishnakumar B
Quoted 1 line

Can't do that at present. Direct mounts are not fully

implemented.

Quoted 5 lines
Quoted 5 lines

My manpage for autofs(5) claims the following:

UNSUPPORTED
This version of the automounter supports direct maps for

FILE, NIS

Quoted 1 line
Quoted 1 line

and LDAP maps only and handles SunOS-style replicated

filesystems

Quoted 6 lines
Quoted 1 line

only to the extent that mount(8) does.

Perhaps I need to change this to be more accurate.

Quoted 2 lines

So is my manpage wrong, or does "direct maps" mean something other

than

Quoted 1 line
Quoted 1 line

"direct mounts"? Or are you referring to the following limitation

(from

Quoted 3 lines
Quoted 3 lines

README.direct v1.3).

NOTE: Due to current design limitations, direct maps will take over

an

Quoted 1 line
Quoted 1 line

entire directory hierarchy. What this means is, if your direct map

key is

Quoted 1 line
Quoted 1 line

/usr/share/bilbo, then /usr will become an automount mount point,

mounting

Quoted 7 lines
Quoted 1 line

over the existing /usr.

I'm refering to the design limitation mentioned in this note.

This will continue to be the case in 4.1.4 (in fact 4.1.x).

I plan on creating a 4.2.0 development branch at some point fairly

soon.

Quoted 8 lines

This will be the top priority for that development.

Ian

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

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

#6 lore raw
In reply to Krishnakumar B
Quoted 8 lines · Krishnakumar B

For example if I have the following setup:

1. I export /export/home from the NFS server, nfs.example.com
2. The NFS server is also configured to use the LDAP server to get automount information, i.e, automount: ldap in /etc/nsswitch.conf
3. auto_home is setup on the LDAP server to spit out something equivalent to:

Quoted 4 lines

* -rw,hard,intr,rsize=8192,wsize=8192,nosuid nfs.example.com:/export/home

Will autofs on nfs.example.com recognize the local mount /export/home and use a loopback mount for /home?

Yes.

Quoted 8 lines

If so, do people recommend using the same
approach (using a direct mount) for /var/spool/mail on nfs.example.com? If not, why not?

Is there any way to disable specify exclude clauses in the mount options to autofs? For example, can I specify some option to autofs so that on a particular machine autofs will not use the automount entry from the directory server.

I don't know a way to specify exclude clauses in the mount options, but you might be able to do something equivalent in the LDAP query itself:
I've written a patch for autofs-4.1.3 to give more flexibility with LDAP requests; in particular, it allows to specify arbitrary search filter expressions, so you should be able to include something along the lines of !(hostname=$MYSELF) into the query, if that is what you need. The patched version of modules/lookup_ldap.c is available at
http://www.timof.qipc.org/autofs-4.1.3-patch

Regards,

Timo Felbinger

Signature · 4 lines
Timo Felbinger                  <[email protected]>
Quantum Physics Group           http://www.quantum.physik.uni-potsdam.de
Institut fuer Physik            Tel: +49 331 977 1793      Fax: -1767
Universitaet Potsdam, Germany
#7 lore raw
In reply to Krishnakumar B
Quoted 11 lines · Krishnakumar B

... However it is not clear how
autofs mounts directories residing on the local filesystem on the NFS server.

For example if I have the following setup:

1. I export /export/home from the NFS server, nfs.example.com
2. The NFS server is also configured to use the LDAP server to get automount information, i.e, automount: ldap in /etc/nsswitch.conf
3. auto_home is setup on the LDAP server to spit out something equivalent to:

Quoted 4 lines

* -rw,hard,intr,rsize=8192,wsize=8192,nosuid nfs.example.com:/export/home

Will autofs on nfs.example.com recognize the local mount /export/home and use a loopback mount for /home?

It works for me. I use NIS, not LDAP, for the map, but it should work the same for any map type. The automount daemon specifically identifies attempted NFS mounts on the local machine and substitutes bind mounts (on Linux).

Quoted 3 lines

If so, do people recommend using the same
approach (using a direct mount) for /var/spool/mail on nfs.example.com? If not, why not?

There's no problem if the mail user agent and delivery agent are both on the machine that has the mailbox, reaching it either by its own name or via a bind/loopback mount. However, if either one is on another host, file locking is not completely effective; most likely there is a short window during which both programs can get the lock. The more common symptom is truncated incoming messages; trashed non-deleted messages are also possible. My most recent test of this was last year on Linux kernel 2.4.21 with kernel NFS; I also tested it a number of years ago on SunOS 3.x.

James F. Carter Voice 310 825 2897 FAX 310 206 6673
UCLA-Mathnet; 6115 MSA; 405 Hilgard Ave.; Los Angeles, CA, USA 90095-1555
Email: [email protected] http://www.math.ucla.edu/~jimc (q.v. for PGP key)