shell script ssh command exit status - shell

Question

In a loop in shell script, I am connecting to various servers and running some commands. For example
#!/bin/bash
FILENAME=$1
cat $FILENAME | while read HOST
do
0</dev/null ssh $HOST 'echo password| sudo -S
echo $HOST
echo $?
pwd
echo $?'
done
Here I am running "echo $HOST" and "pwd" commands and I am getting exit status via "echo $?".
My question is that I want to be able to store the exit status of the commands I run remotely in some variable and then ( based on if the command was success or not) , write a log to a local file.
Any help and code is appreciated.

Answer 1

ssh will exit with the exit code of the remote command. For example:
$ ssh localhost exit 10
$ echo $?
10
So after your ssh command exits, you can simply check $?. You need to make sure that you don't mask your return value. For example, your ssh command finishes up with:
echo $?
This will always return 0. What you probably want is something more like this:
while read HOST; do
echo $HOST
if ssh $HOST 'somecommand' < /dev/null; then
echo SUCCESS
else
echo FAIL
done
You could also write it like this:
while read HOST; do
echo $HOST
if ssh $HOST 'somecommand' < /dev/null
if [ $? -eq 0 ]; then
echo SUCCESS
else
echo FAIL
done

Answer 2

You can assign the exit status to a variable as simple as doing:
variable=$?
Right after the command you are trying to inspect. Do not echo $? before or the new value of $? will be the exit code of echo (usually 0).

Answer 3

An interesting approach would be to retrieve the whole output of each ssh command set in a local variable using backticks, or even seperate with a special charachter (for simplicity say ":") something like:
export MYVAR=`ssh $HOST 'echo -n ${HOSTNAME}\:;pwd'`
after this you can use awk to split MYVAR into your results and continue bash testing.

Answer 4

Perhaps prepare the log file on the other side and pipe it to stdout, like this:
ssh -n user#example.com 'x() { local ret; "$#" >&2; ret=$?; echo "[`date +%Y%m%d-%H%M%S` $ret] $*"; return $ret; };
x true
x false
x sh -c "exit 77";' > local-logfile
Basically just prefix everything on the remote you want to invoke with this x wrapper. It works for conditionals, too, as it does not alter the exit code of a command.
You can easily loop this command.
This example writes into the log something like:
[20141218-174611 0] true
[20141218-174611 1] false
[20141218-174611 77] sh -c exit 77
Of course you can make it better parsable or adapt it to your whishes how the logfile shall look like. Note that the uncatched normal stdout of the remote programs is written to stderr (see the redirection in x()).
If you need a recipe to catch and prepare output of a command for the logfile, here is a copy of such a catcher from https://gist.github.com/hilbix/c53d525f113df77e323d - but yes, this is a bit bigger boilerplate to "Run something in current context of shell, postprocessing stdout+stderr without disturbing return code":
# Redirect lines of stdin/stdout to some other function
# outfn and errfn get following arguments
# "cmd args.." "one line full of output"
: catch outfn errfn cmd args..
catch()
{
local ret o1 o2 tmp
tmp=$(mktemp "catch_XXXXXXX.tmp")
mkfifo "$tmp.out"
mkfifo "$tmp.err"
pipestdinto "$1" "${*:3}" <"$tmp.out" &
o1=$!
pipestdinto "$2" "${*:3}" <"$tmp.err" &
o2=$!
"${#:3}" >"$tmp.out" 2>"$tmp.err"
ret=$?
rm -f "$tmp.out" "$tmp.err" "$tmp"
wait $o1
wait $o2
return $ret
}
: pipestdinto cmd args..
pipestdinto()
{
local x
while read -r x; do "$#" "$x" </dev/null; done
}
STAMP()
{
date +%Y%m%d-%H%M%S
}
# example output function
NOTE()
{
echo "NOTE `STAMP`: $*"
}
ERR()
{
echo "ERR `STAMP`: $*" >&2
}
catch_example()
{
# Example use
catch NOTE ERR find /proc -ls
}
See the second last line for an example (scroll down)