What is the cleanest way to ssh and run multiple commands in Bash?

Question

I already have an ssh agent set up, and I can run commands on an external server in Bash script doing stuff like:

ssh blah_server "ls; pwd;"
         


        

Answer 1

This can also be done as follows. Put your commands in a script, let's name it commands-inc.sh

#!/bin/bash
ls some_folder
./someaction.sh
pwd

Save the file

Now run it on the remote server.

ssh user@remote 'bash -s' < /path/to/commands-inc.sh

Never failed for me.

Answer 2

SSH and Run Multiple Commands in Bash.

Separate commands with semicolons within a string, passed to echo, all piped into the ssh command. For example:

echo "df -k;uname -a" | ssh 192.168.79.134

Pseudo-terminal will not be allocated because stdin is not a terminal.
Filesystem     1K-blocks    Used Available Use% Mounted on
/dev/sda2       18274628 2546476  14799848  15% /
tmpfs             183620      72    183548   1% /dev/shm
/dev/sda1         297485   39074    243051  14% /boot
Linux newserv 2.6.32-431.el6.x86_64 #1 SMP Sun Nov 10 22:19:54 EST 2013 x86_64 x86_64 x86_64 GNU/Linux

Answer 3

Not sure if the cleanest for long commands but certainly the easiest:

ssh user@host "cmd1; cmd2; cmd3"

Answer 4

How about a Bash Here Document:

ssh otherhost << EOF
  ls some_folder; 
  ./someaction.sh 'some params'
  pwd
  ./some_other_action 'other params'
EOF

To avoid the problems mentioned by @Globalz in the comments, you may be able to (depending what you're doing on the remote site) get away with replacing the first line with

ssh otherhost /bin/bash << EOF

Note that you can do variable substitution in the Here document, but you may have to deal with quoting issues. For instance, if you quote the "limit string" (ie. EOF in the above), then you can't do variable substitutions. But without quoting the limit string, variables are substituted. For example, if you have defined $NAME above in your shell script, you could do

ssh otherhost /bin/bash << EOF
touch "/tmp/${NAME}"
EOF

and it would create a file on the destination otherhost with the name of whatever you'd assigned to $NAME. Other rules about shell script quoting also apply, but are too complicated to go into here.

Answer 5

The easiest way to configure your system to use single ssh sessions by default with multiplexing.

This can be done by creating a folder for the sockets:

mkdir ~/.ssh/controlmasters

And then adding the following to your .ssh configuration:

Host *
    ControlMaster auto
    ControlPath ~/.ssh/controlmasters/%r@%h:%p.socket
    ControlMaster auto
    ControlPersist 10m

Now, you do not need to modify any of your code. This allows multiple calls to ssh and scp without creating multiple sessions, which is useful when there needs to be more interaction between your local and remote machines.

Thanks to @terminus's answer, http://www.cyberciti.biz/faq/linux-unix-osx-bsd-ssh-multiplexing-to-speed-up-ssh-connections/ and https://en.wikibooks.org/wiki/OpenSSH/Cookbook/Multiplexing.

Answer 6

For anyone stumbling over here like me - I had success with escaping the semicolon and the newline:

First step: the semicolon. This way, we do not break the ssh command:

ssh <host> echo test\;ls
                    ^ backslash!

Listed the remote hosts /home directory (logged in as root), whereas

ssh <host> echo test;ls
                    ^ NO backslash

listed the current working directory.

Next step: breaking up the line:

                      v another backslash!
ssh <host> echo test\;\
ls

This again listed the remote working directory - improved formatting:

ssh <host>\
  echo test\;\
  ls

If really nicer than here document or quotes around broken lines - well, not me to decide...

(Using bash, Ubuntu 14.04 LTS.)