Your remote shell command will by default execute in the user's $HOME as working directory. If you're giving the full path to the script to execute, that will of course properly launch the remote script, but that script's working directory will still be the user's home. Hence, as config.sh is only a relative file reference, and since it not in the working directory, you cannot source it.
You have several options for solving that:
Before calling the remote script in the sh shell, cd into the directory where it is stored:
ssh <username>@<ip> "cd /path/to/script; sh <script name> <parameter>"
If you can change the remote script, inside the script, at the beginning, cd to its own path so that your working directory changes for all future calls within that script. The working directory of your shell (from which you call the script) will not change if you execute the script with sh.
- In the script, call
config.sh by its absolute path, e.g. source /path/to/config.sh instead of source config.sh. However, that will not be very portable. Or, better:
- In the script, call
config.sh with reference to the current script, e.g. source "$(dirname "$0")/config.sh"
Generally, when scripting, you should try to avoid making relative references if you cannot be sure that the script itself is called from the right working directory. For example, what if a script includes a rm -rf <relative path> statement, but you execute it from the wrong working directory? It could irrevocably delete the wrong data.
