watch file size on linux

Question

I want to watch the growing size of a single file, so i use this command:

texai@maelstrom ~$ ls -lh club_prod.sql | awk '{print $5}'
116M

Now I want to watch that result each 5 seconds so:

texai@maelstrom ~$ watch -n 5 ls -lh club_prod.sql | awk '{print $5}'

but this command doesn't return any result

Using `watch` is not in the same class as parsing within a script... although I admit in my case I would not filter the output of `ls -l` at all when using `watch` with it. — geekosaur, May 11 '12 at 16:23

score 81 · Accepted Answer · answered May 11 '12 at 16:20

81

You're piping the output of watch into awk. If you simplify your command line, what you have is:

 watch <some arguments> | awk '{print $5}'

That's not what you want. Try:

watch -n 5 "ls -lh club_prod.sql | awk '{print \$5}'"

answered May 11 '12 at 16:20

larsks

277,717
41
399
399

watch -n 5 "ls -lh acro | awk '{print \$5}' >> acro_size.txt" works for me, when I do it like above it only shows the size one time? (acro is just my test file) – Emanuel Berg May 11 '12 at 19:05
Hmm, works for me as written. If I change the size of the file, the display updates every 5 seconds. Your example in the comment above appears to be redirecting the output from awk to a file...which means you're hardly benefiting from using `watch`, which is meant as a display tool. – larsks May 11 '12 at 19:23
1

Aha, now I see, it worked for me too (probably), only I expected to get I list, not an update of the entire screen. If you would like to do it to a file as I did, there is probably a better command for just repeating without showing. My mistake. – Emanuel Berg May 12 '12 at 18:31
`while :; do ls -lh club_prod.sql >> acro_size.txt; sleep 60; done` – larsks May 12 '12 at 19:44
2

Why use `ls`/`awk` and piping, when there is a single command `du` for it? – tamasgal Nov 02 '12 at 08:53
I was answering the question of why the `watch` statement wasn't working. There are certainly more effective ways of getting at the size of a file. – larsks Nov 02 '12 at 11:24

score 50 · Answer 2 · answered May 11 '12 at 16:22

50

watch -n 5 "du -h club_prod.sql"

answered May 11 '12 at 16:22

tamasgal

24,826
18
96
135

1

This works well on directory with subdirectories psql dumps with the -s flag: ```watch -n 5 "du -sh directory_with_subdirectories/"``` – pizoelectric Dec 16 '21 at 14:28

Jacajack · Answer 3 · 2018-03-28T10:38:34.617

24

Not exactly related, but if you want to monitor growth rate of some file, you could use following command:

tail -f yourfile.txt | pv > /dev/null

tail -f - outputs data appended to file
pv - measures data flow through pipe
> /dev/null - standard output gets discarded

Note: sometimes pv may be not preinstalled

I hope this will help somebody :)

edited Mar 28 '18 at 10:38

answered May 03 '17 at 11:09

Jacajack

759
2
11
23

score 7 · Answer 4 · answered May 11 '12 at 16:20

You need to quote the pipeline so that it is done within watch.

watch -n 5 "ls -lh club_prod.sql | awk '{print \$5}'"

Note also the \ added to \$5 because the outer quotes are now double quotes, in which $-variables are expanded. (Other methods of quoting are generally uglier than this.)

myki · Answer 5 · 2012-05-11T16:26:32.703

4

watch -n 5 "ls -lh club_prod.sql | awk '{print \$5}'"

edited May 11 '12 at 16:26

answered May 11 '12 at 16:20

myki

773
5
7

Note that `$5` in this example will get eaten by the shell. – larsks May 11 '12 at 16:21
In my case using Ubuntu server I had to do print \$6 not sure if this is because linux now has more output on an ls -lh? But if someone runs into this try 6 ;) – Uncle Iroh Jul 17 '18 at 16:49

score 2 · Answer 6 · answered Jan 22 '18 at 14:21

2

For fast detailed growth watching of a file, every 0.1 second:

watch -n 0.1 "ls -l /mnt/some/file | awk '{print \$5}' | sed -re ' :rep ; s/([0-9])([0-9]{3})($|[^0-9])/\1,\2\3/ ; t rep '"

This will produce something like 62,673,539,072.

answered Jan 22 '18 at 14:21

Tim Chaubet

498
1
7
15

score 2 · Answer 7 · answered Dec 19 '18 at 11:10

The usage of watch is correct, but the usage of ls I would avoid. I would recommend the usage of stat or du, but this depends on what you want.

du: If you want the space occupied on your drive
stat: If you want the number of bytes your file contains (how many bytes can I read from the file)

Imagine working with a compressed file system, or with processing sparse files, internal fragmentation, indirect blocks ...

For both cases, the result would be:

$ watch -n 5 'stat --printf "%s\n" file'
$ watch -n 5 'du -B1 file'

Both results can actually be obtained in a single command with stat:

$ watch -n 5 'stat --printf "%s %b %B\n" file'

The product of the last two columns is the result of du.

score 0 · Answer 8 · edited Dec 19 '18 at 09:59

0

You can perform this like that:

while true; do
  du -s **file_or_directory**
 sleep **time_interval**
done

edited Dec 19 '18 at 09:59

Bsquare ℬℬ

4,423
11
24
44

answered Dec 19 '18 at 09:37

Paresh Solanki

44
3

Marcelo Pacheco · Answer 9 · 2019-08-26T21:46:34.533

0

#!/bin/bash  
# Watch File Size and Growth  
# Author: Marcelo Pacheco - marcelo@m2j.com.br  
# Syntax: watchfilesize filetomonitor
nm="$1"  
while true  
do  
  sz=$(stat -c %s "$nm")  
  sleep 1m  
  sz1=$(stat -c %s "$nm")  
  echo Growth: $(((sz1-sz)/1024))KB/min Size: $((sz1/1024/1024))MB  
  sz=$sz1  
done

edited Aug 26 '19 at 21:46

answered Aug 26 '19 at 19:58

Marcelo Pacheco

152
1
5

watch file size on linux

9 Answers9