If condition is not working in shell scripting

Question

I am monitoring CPU usage and based on that I want to display message. I have written following script:

#!/bin/bash

typeset -F limit1=1.0  # Declare variables as floats.
typeset -F limit2=2.0
echo "-------------------------------------------"
echo "Server_Name   CPU(%)"
echo "-------------------------------------------"
for server in `more /opt/scripts/server-list1.txt`
do
    scpu1=$( ssh $server cat /proc/stat | awk '/cpu/{printf("%.1f\n"),
    ($2+$4)*100/($2+$4+$5)}' |  awk  '{print $0}' | head -1)
    echo "$server   $scpu1"
done | column -t
echo "-------------------------------------------"
for server in `more /opt/scripts/server-list2.txt`
do
    scpu2=$( ssh $server cat /proc/stat | awk '/cpu/{printf("%.2f\n"),
    ($2+$4)*100/($2+$4+$5)}' |  awk '{print $0}' | head -1)
    echo "$server   $scpu2"
done | column -t
echo "-------------------------------------------"
if [[ $scpu1 -ge $limit2 && $scpu2 -le $limit1 ]] ; then
    echo " CPU utilization of node1 is above 2% and node2 is below 1%"
elif [[ $scpu1 -le $limit1  &&  $scpu2 -ge $limit2 ]] ; then
    echo " CPU utilization of node2 is above 2% and node1 is below 1%"
elif [[ $scpu1 -ge $limit2  &&  $scpu2 -ge $limit2 ]] ; then
    echo "CPU utilization of node1 and node2 is above 2% "
else
    echo " Nothing to do"
fi

Though my node1 CPU is less than 2% and node2 CPU is greater than 1% it is giving output as: CPU utilization of node1 is above 2% and node2 is below 1%. Irrespective of condition it is giving the same output as above.

Answer 1

&& is a bash builtin.

However, [ is also a shell builtin that redirects to the test command.

If you look in the manual of test (man test), you can see the following:

EXPRESSION1 -a EXPRESSION2
    both EXPRESSION1 and EXPRESSION2 are true

This means, that you should use -a should be used instead of && inside of [ condition ].

However, as the manual states later:

NOTE: Binary -a and -o are inherently ambiguous.  Use 'test EXPR1  &&  test  EXPR2'  or  'test
EXPR1 || test EXPR2' instead.

This means, the best way of doing this is:

[[ $scpu1 -ge $limit2 ]] && [[ $scpu2 -le $limit1 ]]

As @that other guy points out in the comments of the question:

ShellCheck correctly identifies this problem as a variation of stackoverflow.com/questions/16854280/… . Your variables assignments are forgotten because they happen in subshells.

You assign variables inside for loops, which is ok but the output of the loop is piped to column -t. This pipe creates 2 subshells (for the command before and after the pipe) and the variables of subshells is discarded after the subshell ends.

For example, this happens to scpu1 here:

for server in `more /opt/scripts/server-list1.txt`
do
    scpu1=$( ssh $server cat /proc/stat | awk '/cpu/{printf("%.1f\n"),
    ($2+$4)*100/($2+$4+$5)}' |  awk  '{print $0}' | head -1)
    echo "$server   $scpu1"
done | column -t

As the answer of the linked post of @that other guy suggests, you can use output redirection(which does not create subshells) for this:

for server in `more /opt/scripts/server-list1.txt`
do
    scpu1=$( ssh $server cat /proc/stat | awk '/cpu/{printf("%.1f\n"),
    ($2+$4)*100/($2+$4+$5)}' |  awk  '{print $0}' | head -1)
    echo "$server   $scpu1"
done <<< "$(column -t)"

Note that $() is a newer version of ``

Answer 2

The answer highlight additional problems on top of the issues raised by @dan1st answer, in particular, the usage of | columns

From the comment on typeset -F limit1=1.0 # Declare variables as floats., it implies that the script is expecting floating point math (comparison, etc). bash (like most other sh varients) does NOT support arithmetic or comparison on floating point, only integers.

Consider one of two options: performing all math in integers including the load calculation and the limit1/limit2 values). Alternatively, if it's critical to have more precision (fractions of percent), consider scaling all numbers by a factor of 10 (or 100, if that level of precision is needed).

The '-F' of typeset prevent the display of the function body. It is only useful if the function is printed with typedef -p ..., otherwise it is ignored.

#!/bin/bash


   ### No -F 
typeset limit1=1
typeset limit2=2
   ###
echo "-------------------------------------------"
echo "Server_Name   CPU(%)"
echo "-------------------------------------------"
for server in `more /opt/scripts/server-list1.txt`
do
    scpu1=$( ssh $server cat /proc/stat | awk '/cpu/{printf("%.1f\n"),
    ($2+$4)*100/($2+$4+$5)}' |  awk  '{print $0}' | head -1)
    echo "$server   $scpu1"
done | column -t
echo "-------------------------------------------"
for server in `more /opt/scripts/server-list2.txt`
do
    scpu2=$( ssh $server cat /proc/stat | awk '/cpu/{printf("%.2f\n"),
    ($2+$4)*100/($2+$4+$5)}' |  awk '{print $0}' | head -1)
    echo "$server   $scpu2"
done | column -t
echo "-------------------------------------------"

   ### Calculate load (as integer), and compare
load1=$(printf '%.0f' $scpu1)
load2=$(printf '%.0f' $scpu2)

if [[ $load1 -ge $limit2 && $load2 -le $limit1 ]] ; then
    echo " CPU utilization of node1 is above 2% and node2 is below 1%"
elif [[ $load1 -le $limit1  &&  $load2 -ge $limit2 ]] ; then
    echo " CPU utilization of node2 is above 2% and node1 is below 1%"
elif [[ $load11 -ge $limit2  &&  $load2 -ge $limit2 ]] ; then
    echo "CPU utilization of node1 and node2 is above 2% "
else
    echo " Nothing to do"
fi

Also worth looking into few items:

• Script iterate over list of servers (server-list1.txt), but uses the load of the last server only for the message. If you expect multiple servers to be named, consider averaging, or taking the maximum/minimum value as needed.
• Using more is appropriate for interactive commands, Instead of more /opt/scripts/server-list2.txt, to iterate over list of servers from a file use for server in $(</opt/scripts/server-list2.txt)