pexec

Execute the same command on multiple machines in parallel via SSH.

Install

# Make it executable and symlink to your PATH
chmod +x pexec
sudo ln -s $(pwd)/pexec /usr/local/bin/pexec

Usage

# Single command
pexec -H web1,web2,db1 "uptime"

# Multiple -H flags
pexec -H web1,web2 -H db1,db2 "df -h"

# With user
pexec -H root@web1,deploy@web2 "systemctl status nginx"

# From a hosts file (one host per line)
pexec -f hosts.txt "uname -r"

Session mode (preserving environment)

All -c commands run in a single shell session, so sourced environment variables, aliases, and working directory persist across commands:

# Source an env file, then use its variables
pexec -H web1,web2 -c "source /opt/app/env.sh" -c 'echo $APP_HOME' -c "run_app"

# Login shell (auto-sources ~/.bashrc, ~/.profile)
pexec -H web1,web2 --login -c "conda activate myenv" -c "python train.py"

# Execute a local script on remote hosts
pexec -H web1,web2 -s deploy.sh

# Setup env first, then run a script
pexec -H web1,web2 -c "source /opt/env.sh" -s deploy.sh

# Pipe commands from stdin
echo "uptime && df -h" | pexec -H web1,web2

YAML config file

Define hosts, commands, and settings in a YAML file instead of CLI flags:

pexec --config deploy.yaml

# deploy.yaml
hosts:
  - web1
  - web2
  - root@db1

commands:
  - source /opt/app/env.sh
  - echo $APP_HOME
  - systemctl restart myapp

parallel: 4
timeout: 60
login: true
stream: true
# identity: ~/.ssh/deploy_key
# port: 22
# user: deploy
# script: deploy.sh
# no_strict: false
# json: false

CLI flags override config values, so you can use a config as a base and tweak per-run:

# Use config but add an extra host and override parallelism
pexec --config deploy.yaml -H extra-host -p 2

Default config location: ~/.pexec.yaml (loaded automatically if it exists).

Other features

# Stream output in real-time (for long-running commands)
pexec -H web1,web2 --stream -c "source env.sh" -c "make build"

# JSON output (for scripting)
pexec -H web1,web2 --json "hostname"

# Limit parallelism
pexec -H host1,host2,host3,host4 -p 2 "apt update"

# Custom SSH key and port
pexec -H web1 -i ~/.ssh/mykey --port 2222 "whoami"

Options

Flag	Description
`--config`	YAML config file (default: `~/.pexec.yaml`)
`-c, --cmd`	Command to run (repeatable; all run in one session)
`-s, --script`	Local script file to execute on remote hosts
`--login`	Use a login shell (sources ~/.profile, ~/.bashrc)
`--shell`	Custom shell for sessions (default: `bash -s`)
`-H, --hosts`	Comma-separated hosts (repeatable)
`-f, --hosts-file`	File with hosts, one per line
`-u, --user`	SSH user for all hosts
`-p, --parallel`	Max parallel connections (0=unlimited)
`-t, --timeout`	SSH connect timeout in seconds (default: 30)
`--stream`	Stream output in real-time
`--json`	Output as JSON
`--no-strict`	Disable strict host key checking
`-i, --identity`	SSH private key path
`--port`	SSH port

How session mode works

When you use -c multiple times, pexec joins all commands with newlines and pipes them into bash -s (or bash -ls with --login) over a single SSH connection. This means:

Environment variables set or sourced in earlier -c commands are available in later ones
cd in one command affects subsequent commands
It behaves like typing commands into an interactive shell

Default Hosts File

If no hosts are specified, pexec looks for ~/.pexec_hosts.

zyongye/parallel-exec

README