How to transfer a file to a remote host using parallel + ssh?

Question

I am writing scripts to poll a large number of remote hosts over ssh using gnu parallel. I copy the script to them, run it and get the results. Right now I'm using two bundles parallel + scp to transfer the file, and parallel + ssh to get the result. But I can't figure out how to implement this with one bundle of parallel + ssh, simultaneously transfer the file and get the result.

parallel --tagstring {} -j 25 -q -a file_ip_addr_hosts.csv\  
sshpass -p pass ssh user@{}\  
'cat > file_for_send.sh && source file_for_send.sh && echo result' < file_for_send.sh

It is clear that the file is being created, but it is empty.
Everything works fine without using gnu parallel.

My task is to copy and execute a file on multiple hosts using parallel with one command.

Answer 1

According to a good tradition, I post my own solution. As usual, afterwards, it often turns out to be not simple, but very simple:

parallel –tagstring {} -q -a file_ip_addr_hosts.csv\
sshpass -p pass ssh user@{1} $(<file_for_send.sh)

Of course, there are costs, but they are nonsense. In fact, we are passing the script file for execution, but line by line.
Parallel treats this file as one of the data sources.
Therefore, if, for example, you execute your script in a separate shell
(when you get root access su - root sh -c), we combine all commands through &&, otherwise if you create variables in the process, they will be lost.
I didn't find anything about it in the documentation, although I may have searched badly.
If this construction is allowed through jumpphost, then ‘$(<file_for_send.sh )’ we put it in quotation marks.
There is another point, when passing through jumphost, single quotes in the script (file_for_send.sh ) replace with

‘\'’

Maybe it will help someone.