General

Linux

Simple Parallelism

Sometimes you just want to limit the number of background tasks in a script. Here's an easy way if you don't want to refactor it using parallel:

while read line
do
  some_command --with $line >> log.file &
  while [ $(jobs | wc -l) -gt 50 ]; do 
      sleep 2; 
  done
done < source

The inner while loop will sleep until there are less than 50 tasks in the queue.

Local firewalling

To block an incoming IP address:

iptables -A INPUT -s 187.3.166.114 -j DROP

Or a particular port:

iptables -A INPUT -s 65.55.44.100 -p tcp --destination-port 25 -j DROP

The -s (ip address) is optional in this case.

You can also pipe in data from ipdeny.com, eg:

wget -O- --no-check-certificate  https://www.ipdeny.com/ipblocks/data/countries/ru.zone | \
while read i; do
    iptables -A INPUT -s$i -j DROP
done 

Who logged on last?

Normal people use the last command. You can also use this monstrosity if you are feeling hackish:

perl -we '$recs = ""; while (<>) {$recs .= $_};$uid = 0;foreach (split(/(.{292})/s,$recs)) {next if length($_) == 0;my ($binTime,$line,$host) = $_
=~/(.{4})(.{32})(.{256})/;if (defined $line && $line =~ /\w/) {$line =~ s/\x00+//g;$host =~
s/\x00+//g;printf("%5d %s %8s %s\n",$uid,scalar(gmtime(unpack("I4",$binTime))),$line,$host)}$uid++}print"\n"' < /var/log/lastlog

note - this didn't work the last time I tried it (in 2023)

Soak testing servers

Simple, safe way to raise system load:

NUMCPU=$(grep $'^processor\t*:' /proc/cpuinfo |wc -l)
for ((i=0;i<$NUMCPU;i++));do
  echo 'scale=100000;pi=4*a(1);0' | bc -l &
done ;\
sleep 60; \
killall bc

That brings the cpu load up for 60 seconds by calculating pi to absurd precision, thousands of times. Note the test to see how many threads it should run in parallel.

stress is available on Ubuntu platforms which can also test CPU, memory and disk load.

Somewhat hostile way to load up a system (do NOT run on a live server - can take LA over 1000!):

:(){ :|:& };:

(from http://www.cyberciti.biz/faq/understanding-bash-fork-bomb/)

Using find

The options -delete and -ls are a useful optimisation, saves running '-exec '{}' ls -la \;' and is also a lot faster as it doesn't have to spawn an extra process per file.

Unix Bar Charts

Little perl script to turn a stream of numbers into a pretty realtime barchart:

function baa
{
perl -ne '$min=0; $max=6750; $w=75; use POSIX; $d=ceil(log($max)/log(10)); $w-=$d; $v=$_<$min?0:$_>$max?$max:$_;
$s=$w*$v/($max-$min); $bar=join("", ("*")x$s); $bar=~s/.$/|/ if $v==$max; print sprintf("%${d}d ",$_)."$bar\n";'
}

then, something like this:

(while [ 1 ] ; do
 mailq | tail -1 | awk '{ print $5;}'; sleep 1; done) | baa

will generate this:

1700 *****************
1720 ******************
2000 *********************
2040 *********************
2060 *********************
1640 *****************

Strip last column of text file

To strip the last column of a file, use awk like this:

cat file | awk -F. 'NF{NF--};1'

• -F. — use . as the column separator