Creating remote directories in a bash script via ssh (with error trapping)

Question

I'd like to create directories on a remote VM, but only if they do not exist. What's more, if creating the directory results in an error (exit status != 0) I'd like the script to exit as well, or at least print an error. Is there a more succinct way to do this than what I have below?

UPDATE: Just to clarify, I want to test -d first so mkdir -p won't produce output about the directory already existing.

#!/bin/bash

mydirs=(/var/www/files /var/www/photos /var/www/info)
for d in ${mydirs[@]}; do
   ssh remotehost "test -d $d"
   res=$?
   test $res -ne 0 && { ssh remotehost "mkdir -p $d"; res=$?; }
   test $res -ne 0 && { echo "error during mkdir on remote"; exit 1; }
done

Answer 1

Running ssh in a loop is not efficient. Since the script is not interactive, you could pass the entire script to a remote Bash process on stdin, so that loop will run entirely on the remote server, locally:

ssh remotehost bash << "EOF"
mydirs=(/var/www/files /var/www/photos /var/www/info)
for d in ${mydirs[@]}; do
   test -d $d
   res=$?
   test $res -ne 0 && { mkdir -p $d; res=$?; }
   test $res -ne 0 && { echo "error during mkdir on remote"; exit 1; }
done
EOF

Notice the double-quotes around the here document label, this is to avoid variable interpolation. The entire script is passed to the remote shell literally.

I made only the minimal changes to illustrate the point. Some important improvements are well advised:

• All variables used as command line arguments should be double-quoted: "${mydirs[@]}", "$d", and so on.
• As a comment mentioned, when using mkdir -p, it's unnecessary to test if the directory exists.
• Now that the main operations run in a single process, the pipeline can be simplified.
• As another comment mentioned, consider investing in learning a proper system administration tool such as Puppet, Ansible, Chef, or similar.

Answer 2

Succinct:

#!/bin/bash -e
ssh remotehost mkdir -p /var/www/{files,photos,info}

Failure to create any directory will give non-zero exit for ssh, which causes the outer script to exit with error, because of the -e switch.

mkdir will print suitable error messages if it encounters errors. If you prefer the error message to appear on standard output (like echo would do), you can redirect standard error (aka "filehandle 2") to stdout (filehandle 1) by appending 2>&1:

#!/bin/bash -e
ssh remotehost mkdir -p /var/www/{files,photos,info} 2>&1