Re: Correct way to execute same command in parallel on multiple machines?

[Emil Fihlman]

At the same time the documentation says:

--jobs N

       -j N

       --max-procs N

       -P N     Number of jobslots on each machine. Run up to N jobs in parallel.

                0 means as many as possible. Default is 100% which will run one

                job per CPU core on each machine.

So it should by default run a single job on each machine.

Emil

3.2.2018 23.00 "Ole Tange" <ole@tange.dk> kirjoitti:

On Sat, Feb 3, 2018 at 1:28 PM, Emil Fihlman <emil.fihlman@gmail.com> wrote:

> So I'm experiencing an issue with parallel:
> parallel --tag --nonall --sshloginfile nodelist.txt 'find . -type f | wc -l'
> is ironically executed sequentially on each machine instead of in parallel.
>
> What's the correct way to have parallel execute the command in parallel on
> all machines simultaneously?

From man parallel:

--nonall
           [...] GNU parallel will log into --jobs number of
           computers in parallel and run the job on the computer. -j
adjusts how many
           computers to log into in parallel.

So add -j0.

/Ole