Test if there is a filesystem mounted

[Pedro Cirne]

Hi,

How can I test in a shell script if a filesystem is mounted on directory?

Thks,

Pedro

[Luk Vandenbussche · Accepted solution]

grep $dir /etc/mnttab
if [ $? -eq 0 ]
then
echo mounted
fi

[Pedro Cirne]

Thks

[MarkSyder]

I know you've already closed this, but I like to make commands shorter if I can!

if grep $dir /etc/mnttab
then

etc.

Mark Syder (like the drink but spelt different)

[Dennis Handly]

>Mark Syder: I like to make commands shorter

You should also use grep -q:
if grep -q $dir /etc/mnttab; then

[VK2COT]

Hello,

Note that /etc/mnttab can get corrupt or
have missing record for a specific file system.

Do not rely on /etc/mnttab exclusively.

Here is a little snippet for /bin/sh and /bin/bash script:

#!/bin/sh
fsname="/opt"

if [ "`df $fsname | awk '! /awk/ && ! /Mounted on/ { if ( $NF == "'$fsname'" ) {print}}'`" ]
then
echo "PASS: $fsname mounted"
else
echo "FAIL: $fsname not mounted"
fi

Cheers,

VK2COT

PS. The thread is closed so do not worry about assigning points.

[Dennis Handly]

>VK2COT: Here is a little snippet for /bin/sh

I had a little trouble interpreting your script. First of all, you should have used bdf and not df, otherwise it won't work.

I don't understand why you want to exclude "awk"? ! /awk/

Also, you might want to use awk -v instead of something hard to understand with your stuttering quoting:

M=$(bdf $fsname | awk -v fsname=$fsname '! /Mounted on/ { if ($NF == fsname) print}')
if [ "$M" ]; then

[Dennis Handly]

I forgot to mention in my reply to VK2COT that using bdf to see if something is mounted may be expensive. Doing bdf on a NFS directory may take some time. It may be better to just trust mnttab and use syncer(1M) to keep it up to date, 30 seconds seems ok?

[VK2COT]

Hello again,

Dennis, thanks for the comments.
I used df(1) as I cut-and-pasted from
my home Linux server :)

Of course, bdf(1) is more suitable
for HP-UX.

As far as NFS slowdown is concerned,
one can avoid checking it by using:

bdf -t vxfs
bdf -t hfs
and similar

If NFS is really a problem, then the
server is unresponsive anyway, so the
file system is acting almost like
being unmounted...

Alas, if /etc/mnttab is corrupt,
and if one has bad luck,
waiting 30 seconds for syncer(1) to
update the file can be too long.

# uname -a
HP-UX myserv B.11.11 U 9000/800 1448733058 unlimited-user license

# cat /etc/mnttab

# df -k

# bdf
Filesystem kbytes used avail %used Mounted on
/dev/vg00/lvol3 212992 137904 74560 65% /
/dev/vg00/lvol1 298928 56768 212264 21% /stand
/dev/vg00/lvol8 2621440 1416088 1198128 54% /var
/dev/vg00/lvol10 1048576 3388 981331 0% /var/tmp
/dev/vg00/lvol11 1048576 91944 896903 9% /var/opt/perf
/dev/vg00/lvignite 46612480 33730800 12781064 73% /var/opt/ignite
/dev/vg00/lvol12 1048576 160045 833055 16% /var/opt/OV
/dev/vg00/lvol9 4194304 1012790 2982812 25% /var/adm/crash
/dev/vg00/lvol7 2244608 1401872 836176 63% /usr
/dev/vg00/lvol4 524288 93240 427736 18% /tmp
/dev/vg00/lvol6 3276800 2529776 741224 77% /opt
/dev/vg00/lvol5 524288 5968 514592 1% /home

In other words, one might face a race
condition when tests fail because
/etc/mnttab is empty or corrupt.

Commands bdf(1), and mount(1) update
/etc/mnttab on HP-UX, but df(1) does not.
See real-life example above.

My point was not to trust single commands.
If something is critical double-check.

In my Perl OAT script, for example,
I check file systems in many ways. Once,
I found a very weird server that had
/stand unmounted. II was aksed to come to a specific customer and upgrade their server.
It was the first time ever I saw an HP-UX server running happily without /stand.

It is easy to replicate it (I just did it
on one server):

# umount /stand

# dmesg

Sep 6 21:13
Can't get kernel namelist

That is how I knew something was wrong :)

Cheers,

VK2COT