Parallel SSH, batch and command line oriented.
Basic commands, where output is sent prefixed by the host name :
$ pussh -h host1,host2bis,... uname -a
host1 : Linux host1 3.2.0-4-amd64 #1 SMP Debian 3.2.41-2 x86_64 GNU/Linux
host2bis: Linux host2 3.2.0-3-kirkwood #1 Mon Jul 23 22:36:47 UTC 2012 armv5tel GNU/Linux
...
You can fetch the host list from a file or from stdin :
$ pussh -f servers uname -a
$ fetch-servers | pussh -f - uname -a
Use it with pipes as you would normally, for instance to sort server by rootfs usage :
$ pussh -f servers df -h / |grep /dev |sort -rn -k5
Or collect per-host output, like for inventories :
$ pussh -f servers -o %h-hw.txt lshw
And you can even pipe each SSH session output to a specific command (%h is also available if you whish) :
$ pussh -f servers -o '|grep feature' command >output
Obviously, you have the same features for the SSH sessions input :
$ pussh -f servers -i file command
$ pussh -f servers -i %h.data command
$ pussh -f servers -i 'get-data %h|' command
You may mix -i and -o freely.
More interestingly, you can send and execute a command in one go (but it won't handle the dependencies for you, make sur your executable is self-contained or your host environment is compatible with the target ones) :
$ pussh -f servers -u my-deploy.sh args ...
Most of the time the total time of execution is limited by the SSH connection establishement rate, and this rate is itself limited by the SSH agent. It is recommended to tune your rate for a specific host and stick to it. Exceeding rates generate strange connection errors (such as failed ssh-agent auths, 'no route to host', etc.).
$ alias pussh='pussh -r 50'
Speed is mainly sensitive to network latency. You may first login into one of your remote LAN machine with ssh forwarding (ssh -A) then run 'pussh' from there, the closest possible from its targets. Here is a real benchmark on a Gigabit LAN with OpenSSH 5.x on all servers :
$ time pussh -f servers -r 100 date
...
Total: 201 host(s), 4 second(s)
real 0m4.069s
user 0m7.132s
sys 0m3.140s
Pussh has been an internal tool at Bearstech since roughly 2008. It started with 4 or 5 lines of shell and quickly evolved to its current form. It's still heavily used on a 350+ cluster server, and most of the time reach all of its targets in a few seconds. It is coupled to a cloud management solution to generate a useful list of hosts from very simple descriptions.
- Remote's stdout and stderr ends mixed up in the same stdout stream in the end. Any help to circumvent this is welcome.
- There is no way to use the remote exit status.