autofs in hpux 11.11 mounting shared volumes on netapp filer automatically. Why?

[Sammy_2]

Hpux 11.11 (rp4440 box).

When I do a bdf , I see whole bunch of auto mount points that got mounted by itself. How do I stop it from  automatically happening ?

I believe netapp volumes are world accessible from any server. Now, i get one of these mount point not accessible and it is hanging the bdf command and then finishes after 2 minutes. Also causing slowness (45 sec latency)  in logging in by users.

How do I fix the auto mount point ? I need autofs running though since are using Unix ldap and user home directory automatically gets mounted when user logs in.

bdf: file system inbpnetapp2a:/vol/vol0 mounted on /net/inbpnetapp2a/vol/vol0 is not responding

inbpnetapp4a:/vol/SAP_XIS_diskarch2
                   62914560 4692528 58222032    7% /net/inbpnetapp4a/vol/SAP_XIS_diskarch2
inbpnetapp4a:/vol/ojrfscusers1_clone
                   403701760 394944528 8757232   98% /net/inbpnetapp4a/vol/ojrfscusers1_clone
inbpnetapp4a:/vol/web_enrollment_prod2
                    409600    3872  405728    1% /net/inbpnetapp4a/vol/web_enrollment_prod2
inbpnetapp4a:/vol/dfpm_ojrfscdata2_SV_1229519698_2704
                   1625292800 1553747824 71544976   96% /net/inbpnetapp4a/vol/dfpm_ojrfscdata2_SV_1229519698_2704
inbpnetapp4a:/vol/dfpm_ojrfscusers1_SV_1230046337_3470
                   2147483647 1530533976 616949672   71% /net/inbpnetapp4a/vol/dfpm_ojrfscusers1_SV_1230046337_3470

[Dennis Handly]

I was told long ago to never use /net for the automounter.  This caused ALL of the exported filesystems on the server to be  automounted, even if you only specified one.  Not sure if this is your problem?

But I would assume this would affect more of the server than the client?

The solution was to add specific automount maps for each filesystem.

I've seen NFS hangs like this when I had left some old symlinks to /net/machine that was long since retired.  Whenever I did a lsf in the directory, the stat(2) on that symlink would hang for a while.

In the meantime, if inbpnetapp2a:/vol/vol0 is broken, stop exporting it.

[Sammy_2]

Thanks. I am not mounting any of the /net  nfs file s ystems. Apparently, they get mounted by iteself automagically. I see this on other hpux system as well so wondering if someone else sees the same issue I am having. I am guessing with autofs turned on , it scans the network for available nfs and mounts them.

Dont see this issue with solaris or linux

Also, the vol0 is not broken. It was mounted by itself.

Just wanted to know if these automounts can .be stopped by script or patch.?

Also, I am guessing bdf hanging is due to nfs mounts issue.

[Dennis Handly]

>I am not mounting any of the /net NFS file systems. Apparently, they get mounted by itself automagically.

Then one of your users is doing it.

>with autofs turned on, it scans the network for available NFS and mounts them.

Only if someone accesses those filesystems.

>the vol0 is not broken. It was mounted by itself.

It's broken since it has hung bdf.

>wanted to know if these automounts can be stopped by script or patch?

Some user is causing them to be mounted.  Why do you want to stop them?

[Sammy_2]

Thanks Dennis.

How to I stop users from usiing /net to do soft mount. ? I would much rather they do hard mounts .

[Patrick Wallek]

You don't stop users from doing the mounts.  When the user accesses a file system that is set up to be mounted via AUTOFS, it gets mounted.  This is set up by the system administrator.

If you want to use hard NFS mounts, you (or the system admin) will have to configure these mount points to be mounted via NFS.  This is normally done by adding them to the /etc/fstab file so that they get mounted when the system boots.

[Sammy_2]

Not sure if users are doing it.  When I /net  to one of the volumes and then I do a bdf, it  automatically mounts whole bunch of volumes under /net. So, bdf count  jumps from like 30 to 45. The additonal 15 is just /net mount points.  Odd. I see that on another hpux 11.11 server, same thing.

[Pete Randall]

> How to I stop users from usiing /net to do soft mount. ? I would much rather they do hard mounts .

A normal user (non-root) can not do mounts.  Maybe this has changed since 11.11, but my question would have to be why do you feel the need to eliminate these auto-mounts?  I would suspect that they are being used for something and eliminating them is going to break some application somewhere.

[Dennis Handly]

>Not sure if users are doing it.

Then the applications they are using are doing it.

> When I /net  to one of the volumes and then I do a bdf, it  automatically mounts whole bunch of volumes under /net.

This is exactly what I said.  Using /net is bad.  :-(  (Unless you want all 15 at once. :-)

You need to come up with 15+ new automount maps and tell your applications to use those maps instead of /net.

>The additional 15 is just /net mount points.  Odd. I see that on another hpux 11.11 server, same thing.

Exactly what happens to me, too.

Separate automount maps, good, /net, bad.

[Sammy_2]

Dennis,

If it  happens to you, so I am praying that maybe you have the solution.

The users are not doing (since I rebooted and disabled ssh) and neither is the application. The application does not know anythnig about these mount point on netapp. The app is down. It happens on other HPUX 11.11 server (diffrent app) so I have to say it is the OS. The users dont even know about 10.108.99.30 network.

Immediately After reboot, I did a bdf a few times and saw the mount point in the bottom of my email appear . I umount them and they reappear again a few mins later.

Looking at the syslog.log, I see the OS is again looking for certain volumes and cant find it.  That causes login delay .

Again, i dont think app is doing it. I think, OS is scanning numerous volumes  on the netapp filer which has nothing to do with app on the server. Not sure why ?

Since it can't find it goes on to next one.

Basically, I dont want any /net at all.

Linux  Solairs dont have this issue and they are in similar network. ONly hpux.

How can I do that from OS to stop looking for netapp volumes on the network ? I dont want to completely stop nfs client either.

Thanks

syslog.log

==========

Sep 16 14:08:42 ojrdhes3 automount[5147]: 10.108.99.30:/vol/Helidata2: server not responding: RPC: Timed out
Sep 16 14:08:42 ojrdhes3 syslogd: unknown host loghost4
Sep 16 14:09:12 ojrdhes3 automount[5147]: 10.108.99.30:/vol/FLASH_DUMP: server not responding: RPC: Timed out
Sep 16 14:09:12 ojrdhes3 syslogd: unknown host loghost4
Sep 16 14:09:42 ojrdhes3 automount[5147]: 10.108.99.30:/vol/IKS_ERQ_QA: server not responding: RPC: Timed out
Sep 16 14:09:42 ojrdhes3 syslogd: unknown host loghost4
Sep 16 14:10:12 ojrdhes3 automount[5147]: 10.108.99.30:/vol/dbiftp_dev: server not responding: RPC: Timed out
Sep 16 14:10:12 ojrdhes3 syslogd: unknown host loghost4

Automtically mounted these  after a reboot

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

10.108.99.30:/vol/WORM_iso_50
                   83886080 53171276 30714804   63% /tmp_mnt/net/10.108.99.30/vol/WORM_iso_50
10.108.99.30:/vol/WORM_iso_10
                   3355444 2096728 1258716   62% /tmp_mnt/net/10.108.99.30/vol/WORM_iso_10
10.108.99.30:/vol/web_enrollment2
                    409600     428  409172    0% /tmp_mnt/net/10.108.99.30/vol/web_enrollment2
10.108.99.30:/vol/web_enrollment_prod2
                    409600    3868  405732    1% /tmp_mnt/net/10.108.99.30/vol/web_enrollment_prod2

Sep 16 14:10:43 ojrdhes3 automount[5147]: 10.108.99.30:/vol/TEMP4TEAMSITE: server not responding: RPC: Timed out

[Dennis Handly]

>If it  happens to you, so I am praying that maybe you have the solution.

I mentioned several:

1) Ignore it.  The one I just tried only mounted one extra FS.

2) Add automounter maps for every FS.  Change all the applications to reference the new location.

3) Fix the filer to not get NFS hangs for vol0.

4) If you can't do 3) right away, stop exporting the broken FS.

>The users are not doing and neither is the application. The application does not know anything about these mount point on netapp.

In all my cases there is a direct human cause.  They are referencing /net/...

If you can't find a cause like that, not much you can do except disallow all /net access?

>The users don't even know about 10.108.99.30 network.

All they need to know is: /net/inbpnetapp4a

>I did a bdf a few times and saw the mount point in the bottom of my email appear . I umount them and they reappear 

Some process is referencing those filesystems.  They could be in $PATH variables of the users trying to login.

>I see the OS is again looking for certain volumes and can't find it.  That causes login delay.

Why is that server exporting a broken filesystem, fix it.

>OS is scanning numerous volumes on the netapp filer

It is doing a find(1) in /net/?  Some backup process?

 >I don't want any /net at all.

Then remove that from the automount maps.

>Linux  & Solaris don't have this issue and they are in similar network. Only HP-UX.

Since they aren't set up the same way, this isn't a fair comparison.
 
>How can I do that from OS to stop looking for netapp volumes on the network?

This OS isn't doing for fun.  It is some process you have running.  

automount[5147]: 10.108.99.30:/vol/Helidata2: server not responding: RPC: Timed out
automount[5147]: 10.108.99.30:/vol/FLASH_DUMP: server not responding: RPC: Timed out
automount[5147]: 10.108.99.30:/vol/IKS_ERQ_QA: server not responding: RPC: Timed out
automount[5147]: 10.108.99.30:/vol/dbiftp_dev: server not responding: RPC: Timed out
ojrdhes3 syslogd: unknown host loghost4

Which system is 10.108.99.30?

Who is loghost4?  Any connection?

>Automatically mounted these after a reboot
      /tmp_mnt/net/10.108.99.30/vol/WORM_iso_50
      /tmp_mnt/net/10.108.99.30/vol/WORM_iso_10
      /tmp_mnt/net/10.108.99.30/vol/web_enrollment2
      /tmp_mnt/net/10.108.99.30/vol/web_enrollment_prod2

Did you want them?

>automount[5147]: 10.108.99.30:/vol/TEMP4TEAMSITE: server not responding: RPC: Timed out

Why is this broken and not responding?

[Sammy_2]

Thanks Dennis for answering all of my questions.

How do you disable /net mounts on hPUX 11.11 and still have nfs client running so we can do a hard mount , if needed ?

Thanks

[Dennis Handly]

>How do you disable /net mounts on HP-UX 11.11 and still have NFS client running so we can do a hard mount , if needed?

If you want to hard mount, you edit /etc/fstab.  Or you use the mount(1m) command.

To disable /net, you need to find the autofs command line and edit the configuration file:

UNIX95=EXTENDED_PS ps -H -ef  -x | grep auto

[Matti_Kurkela]

I'd like to expand a bit on Dennis's answer.

HP-UX 11.11 has two NFS automounters: the old-style "automount" and the new "autofs". Both use the same configuration files, but they have separate enable/disable variables in /etc/rc.config.d/nfsconf. Newer HP-UX versions have only autofs.

If you don't need automounter/autofs but need the NFS client, do this:

sh /sbin/init.d/nfs.client stop

vi /etc/rc.config.d/nfsconf
<make sure NFS_CLIENT="1">
<set AUTOMOUNT="0" and AUTOFS="0">

sh /sbin/init.d/nfs.client start

 Now you should have the NFS client running without the automounter. (Note: 11.31 makes stopping the automounter simpler by having its own start/stop script /sbin/init.d/autofs. It's still controlled by the AUTOFS variable in /etc/rc.config.d/nfsconf, though.)

If you need autofs/automounter but don't want /net, comment out the /net definition from the autofs/automounter configuration file /etc/auto_master, then send a "kill -HUP" to the automountd daemon (if you're using the older-style automounter) or run /usr/sbin/automount (if you're using autofs).

[Sammy_2]

Thanks Matti. I think I want automounter but not /net and the solution you provided is probably just what I need.

I will try your and Dennis helpful ideas to get past my problem. Thanks to Dennis and Matti.  I have to figure out how to give pts in this forum.

[Dennis Handly]

>I have to figure out how to give points in this forum.

Just click on the Kudos! star by each reply that helped you.  And set the "Accept as Solution" (on Options for each reply) on the best one.

>sh /sbin/init.d/nfs.client stop

The "sh" isn't needed since each rc(1m) script should start with "#!/sbin/sh".

[Matti_Kurkela]

Just click on the white-on-blue Kudos star to the left of a message if you feel it contains good information.

Apparently you can give Kudos to any posts except your own, even if you haven't started the thread.