monitoring disk space

PTBmilo · 02-20-2003, 02:08 PM

It looks like it's messing up here:

Code:

for i in $( fdisk -l | cut -c1-9 | grep /dev/ )

Did you put a 'l' (as in Lamda) for the argument after fdisk?

If that's not it, post the output of:
bash$ fdisk -l | cut -c1-9 | grep /dev/

I'm thinking that this method might not be very portable....

p.s.- Note my reply to acid_kewpie.... it turns out that 'paste' wanted a file... but we don't need paste if we put it all in an echo containing `date`.

cuss · 02-20-2003, 02:46 PM

Here is the output from the command you suggested:

-------------------------------------------------------------------------
[root@xxxx]# bash$ /sbin/fdisk -l | cut -c1-9 | grep /dev/
bash: bash$: command not found
--------------------------------------------------------------------------

Note that I get the same output from the server that your script works on.

Also, what works for me is '/sbin/fdisk -l'on my one server (that is where fdisk is located (redhat 7.2). In your script you have 'fdisk -l'.

With the changes from yourelf and acid_kewpie would this be the final script (in my case):

-----------------------------------------------------------------------------------
#!/bin/bash
for i in $( /sbin/fdisk -l | cut -c1-9 | grep /dev/ ) # Get all partitions
do
if [ "$( df -h $i | grep -v U | cut -c40-42 )" -ge "80" ]; then

date > /tmp/date_output.txt

echo "`df -h $i | cut -c1-10,40-43 | grep /dev` `date`" \
>> /var/log/disk_free.log
fi
done
------------------------------------------------------------------------------------

Thanks.

PTBmilo · 02-20-2003, 02:58 PM

Congrats... I'm glad it works.

p.s.- I didn't mean for you to put the "bash$" in there

Also, you don't need the 'date > /tmp/date_output.txt' line in there anymore (it's not used because of the alterations mentioned.)

edit:
If you want to see what is going on in here, single out the commands. Start with the first, then add a pipe, then the next, and so on. ex (I'm not showing the output):

Code:

[/home/milo]> /sbin/fdisk -l
[/home/milo]> /sbin/fdisk -l | cut -c1-9
[/home/milo]> /sbin/fdisk -l | cut -c1-9 | grep /dev/

cuss · 02-20-2003, 03:45 PM

Oops my bad. The script works on the server where all partitions are /dev/hda. On the server with 5 /dev/hdc partitions and 1 /dev/sda partition this is the output i get when i run your command.

------------------------------------------------------------------------------------
[root@xxxx]# /sbin/fdisk -l

Usage: fdisk [-l] [-b SSZ] [-u] device
E.g.: fdisk /dev/hda (for the first IDE disk)
or: fdisk /dev/sdc (for the third SCSI disk)
or: fdisk /dev/eda (for the first PS/2 ESDI drive)
or: fdisk /dev/rd/c0d0 or: fdisk /dev/ida/c0d0 (for RAID devices)
...
------------------------------------------------------------------------------------

And this would be the good copy of the script:

-----------------------------------------------------------------------------------
#!/bin/bash
for i in $( /sbin/fdisk -l | cut -c1-9 | grep /dev/ ) # Get all partitions
do

if [ "$( df -h $i | grep -v U | cut -c40-42 )" -ge "80" ]; then

echo "`df -h $i | cut -c1-10,40-43 | grep /dev` `date`" \
>> /var/log/disk_free.log

fi

done
------------------------------------------------------------------------------------

Thanks!

PTBmilo · 02-20-2003, 04:04 PM

What does this give you (on the server that the script doesn't work on)?

(first one will give the fdisk version)

Code:

[/home/milo]> fdisk -v
[/home/milo]> fdisk /dev/[hs]d? -l

What distro is it on?

cuss · 02-21-2003, 10:01 AM

I try both of the above commands. Here is the output:

[root@xxxxx /root]# /sbin/fdisk -v
fdisk v2.9w

[root@xxxxx /root]# /sbin/fdisk /dev/[hs]d? -l
Device Boot Start End Blocks Id System
#All information for /dev/hdc and /dev/sda is listed here

It is on a redhat system, kernel 2.2.12-20. I believe redhat version 6.1.

acid_kewpie · 02-21-2003, 10:39 AM

yeah some versions of fdisk insist on being told which device to list, some default to all.... not sure if it's newer or older versions that do this...

PTBmilo · 02-21-2003, 10:42 AM

Cool, if that lists everything, then I thin you can stick that into the program.

Code:

for i in $( /sbin/fdisk /dev/[hs]d? -l | cut -c1-9 | grep /dev/ )

If something screws up, then post exactly what the output of just the fdisk command is.

By the way, the [hs] means it will match everything that starts with an h or an s, and the ? will match any single character.

cuss · 02-21-2003, 03:28 PM

Well, I've got the script on 3 servers now and all produce output to /var/log/disk_free.log every 60 minutes which is great. The problem now is that i setup cron to email me if disk space is over a certain percentage once it starts the script every 60 minutes. However the emails aren't getting to me (one of the servers is over its threshold so i should get something). Don't know if you know much about cron. But thanks for the help on the disk space script!!! Here is my crontab entry:

------------------------------------------------------------------------------
#check disk space every 60 minutes
MAILTO=me@mycompany.com
00 * * * * /usr/local/bin/disk_space.sh
------------------------------------------------------------------------------

I was receving the emails yesterday telling me i had my syntax for the /sbin/fdisk line incomplete. Any suggestions?

Thanks.

PTBmilo · 02-24-2003, 05:32 PM

You can use the script to notify you. I don't really know the format that you want in the e-mails, but here's an example to work from:

below the ">> var/log/disk_free.log" line in the script:

Code:

tail /var/log/disk_free.log | mail -s "DISK USAGE ALERT ON `hostname -f`" your@address

The only thing about this is that it will send you 1 email for every single partition that is over 80% used (every hour). Maybe if I get some time I'll post a better solution.

EDIT:
What do you mean you were getting e-mails saying that the fdisk command was incomplete? What did they say? I can try to help if you tell me what is going wrong.

rnturn · 02-24-2003, 07:15 PM

Quote:

Originally posted by PTBmilo

The only thing about this is that it will send you 1 email for every single partition that is over 80% used (every hour). Maybe if I get some time I'll post a better solution.

In cases like that, I use something like:

Code:

#!/bin/bash

REPORT=/tmp/df_report_$$.txt
THRESH=80
TRUE="1 eq 1"
FALSE="1 eq 0"

WROTETOREPORT=$FALSE

df -k | grep -v "Mounted on" | while read DFREC
do
    SPECIAL=`echo $DFREC | awk '{ print $1 }'`
    FS=`echo $DFREC | awk '{ print $6 }'`
    USE=`echo $DFREC | awk '{ print $5 }'`
    PERCENT=`echo $USE | sed 's/%//'`
    if [ $PERCENT -gt $THRESH ]
    then
        WROTETOREPORT=$TRUE
        echo "$SPECIAL ($FS) is at $PERCENT percent." >> $REPORT
    fi
done
if [ $WROTETOREPORT ]
then
    cat $REPORT | mutt -s "Filesystems above $THRESH percent on date +'%Y-%m-%d %H:%M'" root@localhost
    rm -f $REPORT
fi

Or something like that. It looks wierd at first but I use the variables (which are normally sourced, loaded as a module, or included as a header) to hold the boolean values so I don't have to remember what's true/false in shell, what's true/false in perl, etc.

cuss · 02-25-2003, 12:44 PM

Hi PTPmilo,

Thanks again for all of your responses and most recently your one line solution to my emailing problem.....it's no longer a problem. All 3 servers are doing what they should be now.

----------------------------------------------------------------------------------
EDIT:
What do you mean you were getting e-mails saying that the fdisk command was incomplete? What did they say? I can try to help if you tell me what is going wrong.
-----------------------------------------------------------------------------------
No longer an issue.....thanks anyways.

Thanks to you to rnturn for the script you provided and everyone else for their help along the way. Hopefully this issue is now a closed case (at least by me).

Cheers.