Skip to contentSkip to conversation

autofs-4.1.3 not working properly

DISCUSSION Ian Kent 2 participants 5 messages Nov 27, 2005 – Nov 30, 2005 1 message not in the archive Kernel automounter

lore
Share & tools
lore

Ian Kent

author maintainer
#1 lore raw
In reply to a message not in the archive
Quoted 35 lines · Prakash Velayutham
Quoted 1 line
Quoted 1 line
Quoted 1 line

Ian Kent <[email protected]> 11/25/05 1:47 PM >>>

Quoted 5 lines · Prakash Velayutham

Hi,

I am new to this list, so please forgive my ignorances.
I had a SuSE Pro 9.0 system running autofs (v3) running earlier. The autofs itself did not have any issues at all until I decided to

upgrade

Quoted 1 line

the system to SuSE 9.3. It was a clean install, and autofs4-4.1.3

became

Quoted 8 lines

the default kernel autofs module. My autofs master map comes from a OpenLDAP server and it contains 3 different mount maps. /users (LDAP map)
/protein/users (LDAP map)
/import/users (LDAP map)
I also have a file-based map in this server (/export/users).

Recently I was trying to move a user's home dir from server1 to

server2.

Quoted 11 lines

After moving his home dir and making the relevant changes to his LDAP entry (homeDirectory attribute), I tried to restart autofs in the above-mentioned server. The server already had several users logged in

under /protein/users. Though the restart did not complain, I noticed that autofs status showed "Configured mount points" correctly and removed the currently mounted mount points from "Active mount points".

Is there a reason why? Also strangely the ownerships of the previously

mounted dirs had been changed to root:root.

I'm not sure what is not working or what has been broken. What is the actual problem and symptom?

Quoted 6 lines

Ian

Thanks Ian for a reply. What if I restart autofs when a user whose home dir is mounted through autofs is already logged into the system (and hence at least one of the automount entries is being used)? What will the system do in that case?

On runing "reload" it should, depending on version and patch levels re-read and update the map, leave the mounted directory mounted and leave the stale map entry for cleanup next time the map is reloaded and the entry isn't mounted.

"Restart"ing is much more agressive and I wouldn't recommend it if you have mount that are in use. To restart you really need to have nothing actually in use.

Quoted 5 lines

And also if I change the ldap attribute "homeDirectory" for a user, do I have to restart autofs in a system for that change to be seen. Because I sometimes see that the system has cached the user's attributes from LDAP and tries to use that and fails.

autofs doesn't use that attribute so no, but you'll need to be sure that the automount map entry that is used to access that directory is still valid following the change and if it also had to be changed then you might need to "reload" autofs. It's worth pointing out that later versions (most RedHat versions and 4.1.4 I think) of autofs should recognise this change on access without needing to re-load the map.

The other thing I noticed about your query was the question about the root owned directory. At variuos times in the past development autofs has been (mostly intentionally) lazy about cleaning up mount point directories. When autofs directories don't have a filesystem mounted on them they will appear root owned. It shouldn't make a difference to operation.

Ian

#2 lore raw
Quoted 59 lines · Ian Kent
Quoted 37 lines · Prakash Velayutham
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

Ian Kent <[email protected]> 11/25/05 1:47 PM >>>

Quoted 5 lines · Prakash Velayutham

Hi,

I am new to this list, so please forgive my ignorances.
I had a SuSE Pro 9.0 system running autofs (v3) running earlier. The autofs itself did not have any issues at all until I decided to

upgrade

Quoted 2 lines

the system to SuSE 9.3. It was a clean install, and autofs4-4.1.3

became

Quoted 8 lines

the default kernel autofs module. My autofs master map comes from a OpenLDAP server and it contains 3 different mount maps. /users (LDAP map)
/protein/users (LDAP map)
/import/users (LDAP map)
I also have a file-based map in this server (/export/users).

Recently I was trying to move a user's home dir from server1 to

server2.

Quoted 8 lines

After moving his home dir and making the relevant changes to his LDAP entry (homeDirectory attribute), I tried to restart autofs in the above-mentioned server. The server already had several users logged in

under /protein/users. Though the restart did not complain, I noticed that autofs status showed "Configured mount points" correctly and removed the currently mounted mount points from "Active mount points".

Is there a reason why? Also strangely the ownerships of the previously

mounted dirs had been changed to root:root.

I'm not sure what is not working or what has been broken. What is the actual problem and symptom?

Quoted 7 lines

Ian

Thanks Ian for a reply. What if I restart autofs when a user whose home dir is mounted through autofs is already logged into the system (and hence at least one of the automount entries is being used)? What will the system do in that case?

On runing "reload" it should, depending on version and patch levels re-read and update the map, leave the mounted directory mounted and leave the stale map entry for cleanup next time the map is reloaded and the entry isn't mounted.

"Restart"ing is much more agressive and I wouldn't recommend it if you have mount that are in use. To restart you really need to have nothing actually in use.

Quoted 5 lines

And also if I change the ldap attribute "homeDirectory" for a user, do I have to restart autofs in a system for that change to be seen. Because I sometimes see that the system has cached the user's attributes from LDAP and tries to use that and fails.

Quoted 5 lines

autofs doesn't use that attribute so no, but you'll need to be sure that the automount map entry that is used to access that directory is still valid following the change and if it also had to be changed then you might need to "reload" autofs. It's worth pointing out that later versions (most RedHat versions and 4.1.4 I think) of autofs should recognise this change on access without needing to re-load the map.

The other thing I noticed about your query was the question about the root owned directory. At variuos times in the past development autofs has been (mostly intentionally) lazy about cleaning up mount point directories. When autofs directories don't have a filesystem mounted on them they will appear root owned. It shouldn't make a difference to operation.

Ian

Thanks Ian. I tried out autofs reload and there seems to be a small glitch. It says "checking for changes in auto.master" and then stops some of the automount daemons. But stays there forever. If I do Ctrl-C, then it immediately shows "Starting XXX" for the different automount daemons (except 1 for some reason). Is there a reason for this behaviour. The system I am trying out is SuSE Linux 10.0, autofs-4.1.4-6 with automount master map and the entries coming from a OpenLDAP server.

Thanks,
Prakash

author maintainer
#3 lore raw
Quoted 79 lines · Prakash Velayutham
Quoted 78 lines · Ian Kent
Quoted 39 lines · Prakash Velayutham
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

Ian Kent <[email protected]> 11/25/05 1:47 PM >>>

Quoted 6 lines · Prakash Velayutham

Hi,

I am new to this list, so please forgive my ignorances.
I had a SuSE Pro 9.0 system running autofs (v3) running earlier. The autofs itself did not have any issues at all until I decided to

upgrade

Quoted 2 lines

the system to SuSE 9.3. It was a clean install, and autofs4-4.1.3

became

Quoted 9 lines

the default kernel autofs module. My autofs master map comes from a OpenLDAP server and it contains 3 different mount maps. /users (LDAP map)
/protein/users (LDAP map)
/import/users (LDAP map)
I also have a file-based map in this server (/export/users).

Recently I was trying to move a user's home dir from server1 to

server2.

Quoted 11 lines

After moving his home dir and making the relevant changes to his LDAP entry (homeDirectory attribute), I tried to restart autofs in the above-mentioned server. The server already had several users logged in
under /protein/users. Though the restart did not complain, I
noticed that autofs status showed "Configured mount points" correctly and removed the currently mounted mount points from "Active mount points".
Is there a reason why? Also strangely the ownerships of the
previously
mounted dirs had been changed to root:root.

I'm not sure what is not working or what has been broken. What is the actual problem and symptom?

Quoted 7 lines

Ian

Thanks Ian for a reply. What if I restart autofs when a user whose home dir is mounted through autofs is already logged into the system (and hence at least one of the automount entries is being used)? What will the system do in that case?

On runing "reload" it should, depending on version and patch levels re-read and update the map, leave the mounted directory mounted and leave the stale map entry for cleanup next time the map is reloaded and the entry isn't mounted.
"Restart"ing is much more agressive and I wouldn't recommend it if you have mount that are in use. To restart you really need to have nothing actually in use.

Quoted 5 lines

And also if I change the ldap attribute "homeDirectory" for a user, do I have to restart autofs in a system for that change to be seen. Because I sometimes see that the system has cached the user's attributes from LDAP and tries to use that and fails.

autofs doesn't use that attribute so no, but you'll need to be sure that the automount map entry that is used to access that directory is still valid following the change and if it also had to be changed then you might need to "reload" autofs. It's worth pointing out that later versions (most RedHat versions and 4.1.4 I think) of autofs should recognise this change on access without needing to re-load the map.

The other thing I noticed about your query was the question about the root owned directory. At variuos times in the past development autofs has been (mostly intentionally) lazy about cleaning up mount point directories. When autofs directories don't have a filesystem mounted on them they will appear root owned. It shouldn't make a difference to operation.

Quoted 8 lines
Quoted 1 line

Ian

Thanks Ian. I tried out autofs reload and there seems to be a small glitch. It says "checking for changes in auto.master" and then stops some of the automount daemons. But stays there forever. If I do Ctrl-C, then it immediately shows "Starting XXX" for the different automount daemons (except 1 for some reason). Is there a reason for this behaviour. The system I am trying out is SuSE Linux 10.0, autofs-4.1.4-6 with automount master map and the entries coming from a OpenLDAP server.

Sounds like a bug to me.

I'll try to duplcate it.
Can you offer any additional information on your setup?

Ian

#4 lore raw
Quoted 104 lines · Ian Kent
Quoted 92 lines · Prakash Velayutham
Quoted 82 lines · Ian Kent
Quoted 48 lines · Prakash Velayutham
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

Ian Kent <[email protected]> 11/25/05 1:47 PM >>>

Quoted 7 lines · Prakash Velayutham

Hi,

I am new to this list, so please forgive my ignorances.
I had a SuSE Pro 9.0 system running autofs (v3) running earlier. The autofs itself did not have any issues at all until I decided to

upgrade

Quoted 3 lines

the system to SuSE 9.3. It was a clean install, and autofs4-4.1.3

became

Quoted 10 lines

the default kernel autofs module. My autofs master map comes from a OpenLDAP server and it contains 3 different mount maps. /users (LDAP map)
/protein/users (LDAP map)
/import/users (LDAP map)
I also have a file-based map in this server (/export/users).

Recently I was trying to move a user's home dir from server1 to

server2.

Quoted 12 lines

After moving his home dir and making the relevant changes to his LDAP entry (homeDirectory attribute), I tried to restart autofs in the above-mentioned server. The server already had several users logged in
under /protein/users. Though the restart did not complain, I
noticed that autofs status showed "Configured mount points" correctly and removed the currently mounted mount points from "Active mount points".
Is there a reason why? Also strangely the ownerships of the
previously
mounted dirs had been changed to root:root.

I'm not sure what is not working or what has been broken. What is the actual problem and symptom?

Quoted 8 lines

Ian

Thanks Ian for a reply. What if I restart autofs when a user whose home dir is mounted through autofs is already logged into the system (and hence at least one of the automount entries is being used)? What will the system do in that case?

On runing "reload" it should, depending on version and patch levels re-read and update the map, leave the mounted directory mounted and leave the stale map entry for cleanup next time the map is reloaded and the entry isn't mounted. "Restart"ing is much more agressive and I wouldn't recommend it if you have mount that are in use. To restart you really need to have nothing actually in use.

Quoted 6 lines

And also if I change the ldap attribute "homeDirectory" for a user, do I have to restart autofs in a system for that change to be seen. Because I sometimes see that the system has cached the user's attributes from LDAP and tries to use that and fails.

autofs doesn't use that attribute so no, but you'll need to be sure that the automount map entry that is used to access that directory is still valid following the change and if it also had to be changed then you might need to "reload" autofs. It's worth pointing out that later versions (most RedHat versions and 4.1.4 I think) of autofs should recognise this change on access without needing to re-load the map.

Quoted 8 lines

The other thing I noticed about your query was the question about the root owned directory. At variuos times in the past development autofs has been (mostly intentionally) lazy about cleaning up mount point directories. When autofs directories don't have a filesystem mounted on them they will appear root owned. It shouldn't make a difference to operation.

Ian

Quoted 8 lines

Thanks Ian. I tried out autofs reload and there seems to be a small glitch. It says "checking for changes in auto.master" and then stops some of the automount daemons. But stays there forever. If I do Ctrl-C, then it immediately shows "Starting XXX" for the different automount daemons (except 1 for some reason). Is there a reason for this behaviour. The system I am trying out is SuSE Linux 10.0, autofs-4.1.4-6 with automount master map and the entries coming from a OpenLDAP server.

Quoted 6 lines

Sounds like a bug to me.

I'll try to duplcate it.
Can you offer any additional information on your setup?

Ian

Hi Ian,

Sorry for this long posting. Just giving all the info required. Let me know if you need anything else in particular.

Client details:
SuSE Linux 10.0 (the Open SuSE OS).
autofs-4.1.4-6

auto.master map comes from LDAP server (NIS attributes, not automount attributes) as well as from the file /etc/auto.master (nsswitch.conf has automount: files nis ldap) DN is:
dn: nisMapName=auto.master,o=x,c=x
nisMapName: auto.master
objectClass: nisMap
objectClass: top

4 different map entries (3 through LDAP and 1 through file). DNs are:
dn: nisMapName=auto.users,o=x,c=x
nisMapName: auto.users
objectClass: nisMap
objectClass: top
structuralObjectClass: nisMap

dn: cn=/,nisMapName=auto.users,o=x,c=x
cn: /
nisMapName: auto.users
objectClass: nisObject
structuralObjectClass: nisObject
nisMapEntry: -fstype=nfs,hard,intr x.x.x:/home/&

dn: cn=/users,nisMapName=auto.master,o=x,c=x
cn: /users
nisMapName: auto.master
objectClass: nisObject
nisMapEntry: ldap x.x.x:nisMapName=auto.users,o=x,c=x

dn: cn=/import/users,nisMapName=auto.master,o=x,c=x
cn: /import/users
nisMapName: auto.master
objectClass: nisObject
nisMapEntry: ldap x.x.x:nisMapName=auto.import,o=x,c=x
structuralObjectClass: nisObject

dn: nisMapName=auto.import,o=x,c=x
nisMapName: auto.import
objectClass: nisMap
objectClass: top
structuralObjectClass: nisMap

dn: cn=/,nisMapName=auto.import,o=x,c=x
cn: /
nisMapName: auto.import
objectClass: nisObject
nisMapEntry: -fstype=nfs,hard,intr x.x.x.x:/home/&
structuralObjectClass: nisObject

dn: nisMapName=auto.proteinusers,o=x,c=x
nisMapName: auto.proteinusers
objectClass: nisMap
objectClass: top
structuralObjectClass: nisMap

dn: cn=/,nisMapName=auto.proteinusers,o=x,c=x
cn: /
nisMapName: auto.proteinusers
objectClass: nisObject
nisMapEntry: -fstype=nfs,hard,intr x.x.x.x:/home/&
structuralObjectClass: nisObject

dn: cn=/protein/users,nisMapName=auto.master,o=x,c=x
cn: /protein/users
nisMapName: auto.master
objectClass: nisObject
nisMapEntry: ldap x.x.x:nisMapName=auto.proteinusers,o=x,c=x
structuralObjectClass: nisObject

The file based auto.master has
/export/users auto_users -rw,bg,hard,rsize=8192,wsize=8192
+

Contents of /etc/auto_users:
* x.x.x.x:/pi_home/home/&/.linux

Any help is greatly appreciated.

Thanks,
Prakash

author maintainer
#5 lore raw
Quoted 11 lines · Prakash Velayutham
Quoted 7 lines

Sounds like a bug to me.

I'll try to duplcate it.
Can you offer any additional information on your setup?

Ian

Hi Ian,

Sorry for this long posting. Just giving all the info required. Let me know if you need anything else in particular.

That's OK.

I always have trouble finding/getting hold of the SuSE source rpm to test this stuff out. If you are able to forward one that would be great. Otherwise I'll test with my 4.1.4 and the minimum patches I recommend. It's likey to show up the problem any way but I thought I'd mention it.

Quoted 5 lines

Client details:
SuSE Linux 10.0 (the Open SuSE OS).
autofs-4.1.4-6

This is exactly what I need to duplicate the problem. My test setup is not greatly different anyway. I'll setup and test soon as I get a chance.

Thanks
Ian