How to deploy a shell script to private linux vm - linux

Question

I have a private Azure Linux VM means it can be accessed only from jumpbox(access) vm. I need to deploy a script to this private VM. As this VM cannot even access any storage account/repo, I can't use Custom Script Extension for script deployment. So I thought of deploying the script using az vm run-command invoke by converting the SomeScript.sh to strings and echo it to the virtual machine. Different pieces of my code as below:
SomeScript.sh
#!/bin/bash
#
# CommandToExecute: ./SomeScript.sh ${CUST_NO}
#
#some more code
Function that converts the .sh file to strings:
function getCommandToExecute()
{
local scriptName=$1
local commandToExecute
local currentLocation=$(dirname "$0")
local scriptFullPath="$currentLocation/Environment/VmScripts/$scriptName"
mapfile < $scriptFullPath
printf -v escapedContents "%q\n" "${MAPFILE[#]}"
commandToExecute+="echo "$escapedContents" > /usr/myapps/$scriptName"
echo "$commandToExecute"
}
vm run command:
az vm run-command invoke -g $resourceGroupName \
-n $vmName --command-id RunShellScript \
--scripts "#!/bin/bash\n ${commandToExecute}"
If I use the "#!/bin/bash\n ${commandToExecute}" part (commandToExecute replaced with string scripts) in the RunCommand window in azure portal, the script works fine, but I can't make it work via run command due to this exception:
\n[stdout]\n\n[stderr]\n/bin/sh: 1: /var/lib/waagent/run-command/download/133/script.sh: not found\n"
Any idea what is missing here? Or if there is a better alternative to handle this.

Answer 1

I think quoting the whole script and its deployment script for use with --scripts is a lot of work and error prone too. Luckily, there are easier alternatives for both quoting steps. The documentation of az vm run-command invoke --scripts states
Use #{file} to load script from a file.
Therefore, you can do
deploymentScript() {
echo '#! /usr/bin/env bash'
printf 'tail -n+4 "${BASH_SOURCE[0]}" > %q\n' "$2"
echo 'exit'
cat "$1"
}
deploymentScript local.sh remote.sh > tmpDeploy.sh
az vm run-command invoke ... --scripts '#{tmpDeploy.sh}'
rm tmpDeploy.sh
Replace local.sh with the path of the local script you want to deploy and replace remote.sh with the remote path where the script should be deployed.
If you are lucky, then you might not even need tmpDeploy.sh. Try
az vm ... --scripts "#{<(deploymentScript local.sh remote.sh)}"
Some notes on the implementation:
The deployed script is an exact copy of your script file. Even embedded binary data is kept. The trick with tail $BASH_SOURCE is inspired by this answer.
The script can be arbitrarily long. Even huge scripts won't run into the error Argument list too long imposed by getconf ARG_MAX.

Answer 2

There was issue with escaped string I mentioned above. After correcting the code, it all good now. My final code:
function getCommandToExecute()
{
local scriptName=$1
local commandToExecute
local currentLocation=$(dirname "$0")
local scriptFullPath="$currentLocation/Environment/VmScripts/$scriptName"
local singleStringCommand=""
mapfile -t < $scriptFullPath
for line in "${MAPFILE[#]}";
do
singleStringCommand+="$(printf '%s' "$line" | sed 's/[\"$]/\\&/g')"
singleStringCommand+="\n"
done
commandToExecute+="echo "\"$singleStringCommand\"" > /usr/local/bin/$scriptName;"
echo "$commandToExecute"
}