fuel-pdsh 1.0.0.dev21

Simple pure Python alternative to PDSH for OpenStack Fuel node.

fuel-pdsh is a small utility which allows you to execute one command on the multiple OpenStack nodes from master node (aka Fuel node) remotely using SSH.

Basically it is just a pure Python replacement for PDSH which uses Nailgun API to discover nodes. Also it supports pdcp utility for copying files to multiple hosts in parallel.

fuel-pdsh package contains 2 commandline utilities, fuel-pdsh for executing SSH command on the group of hosts and fuel-pdcp for file copying.

Installation

Before install fuel-pdsh on master node, do the following:

$ yum install python-devel python-pip gcc

If you want system installation, do:

$ pip install fuel-pdsh

If you want to use virtualenv:

$ pip install virtualenv
$ virtualenv -p python2.6 --system-site-packages ~/.fuelpdsh-venv
$ source ~/.fuelpdsh-venv/bin/activate
$ pip install fuel-pdsh

Commandline options

The both of fuel-pdsh and fuel-pdcp have the same options set but different arguments.

usage: fuel-pdsh [-h] [--concurrency CONCURRENCY] [-c CLUSTER_ID]
                 [-w NODE_IDS] [-i IPS] [-n NAME] [-s STATUS] [-g GROUP_ID]
                 [-r ROLES] [-v | -d]
                 command [command ...]

positional arguments:
  command               Command to execute

optional arguments:
  -h, --help            show this help message and exit
  --concurrency CONCURRENCY
                        How many simultaneous connections should be
                        established. By default (4), we are trying to connect
                        to all nodes, no limits.
  -c CLUSTER_ID, --cluster-id CLUSTER_ID
                        Select only nodes which belong to cluster with such
                        ID.
  -w NODE_IDS, --node-ids NODE_IDS
                        Plain comma-separated list of nodes.
  -i IPS, --ips IPS     Plain comma-separated list of node IPs.
  -n NAME, --name NAME  Regular expression for the node name.
  -s STATUS, --status STATUS
                        Node status.
  -g GROUP_ID, --group-id GROUP_ID
                        Group ID.
  -r ROLES, --roles ROLES
                        Node roles.
  -v, --verbose         Be verbose.
  -d, --debug           Be event more verbose, for debugging.

Please contact Sergey Arkhipov <[e-mail]> for issues.

--concurrency

This flag defines how may hosts would be accessed in parallel. If you have 40 nodes but --concurrency is set to 4, only 4 hosts will be managed in parallel. Set 0 if you want to connect all hosts in parallel.

In reality, please do not set this setting to high. fuel-pdsh uses Paramiko for host access and it has well-known issues on slowing down with multiple simultaneous connections. You may have serious performance decrease if you connect to all hosts instead of some limit.

--verbose

This flag makes utilities more verbose. By default, if you execute fuel-pdsh, you will see such output:

$ fuel-pdsh -n contr -- echo hello world
node-4 :  hello world
node-39:  hello world
node-3 :  hello world

If you enable -v, you will get something like this:

$ fuel-pdsh -n contr -v -- echo hello world
*** 140489797273344 >>> Found 3 suitable nodes
*** 140489797273344 >>> Execute with 4 threads
node-4 :  hello world
node-39:  hello world
node-3 :  hello world

So, more verbose, to understand what is going.

--debug

Enables maximal verbosity. Basically, if you are not me, you do not need this level of verbosity. But I need when I debug.

So, if you met with some problems and want to issue a bug, execute utilities with -d and send me an output.

-c, --cluster-id

Defines optional cluster ID for additional node filtering. If no cluster ID is set, utilities will work over all accessible clusters.

$ fuel-pdsh -c 1 -- ls

This will do ls on all nodes in cluster with ID 1.

-w, --node-ids

The most simple selector, just select all nodes by given IDs. So if you want t``-c``, o run a command on nodes with IDs 2, 4 and 8, just pass them as a comma-separated list.

$ fuel-pdsh -w 2,4,8 -- ls

Also, you may prefix them with node- prefix (as you SSH to them). The following command is the same as previous:

$ fuel-pdsh -w node-2,4,node-8 -- ls

-i, --ips

Select only those nodes which have these IPs.

$ fuel-pdsh -i 10.0.0.1,10.0.0.2 -- ls

-n, --name

Filters on the node names. This parameter is just a regular expression for the node name, so there is not point to enter the whole name, just pass a part.

$ fuel-pdsh -n contro -- ls

-s, --status

Filter nodes on their statuses.

$ fuel-pdsh -s ready -- ls

This will ls on all nodes which have status ready.

-g, --group-id

Filters nodes on their group ID.

$ fuel-pdsh -g 10 -- ls

-r, -roles

Filter nodes on their roles.

$ fuel-pdsh -r compute -- ls

fuel-pdsh

fuel-pdsh is a tool to execute commands in parallel on different hosts. Let’s assume you want to restart Apache on all controllers. Then do following:

$ fuel-pdsh -r controller service apache2 restart

Sometimes you need to pass arguments to the command which may be recognized as an arguments for fuel-pdsh itself. No worries, good old -- is supported.

$ fuel-pdsh -r controller -- manage.py --noinput

Sometimes you have to invoke several commands. No worries again:

$ fuel-pdsh -r controller -- sh -c "command1 && command2; command3"

fuel-pdcp

fuel-pdcp is a utility to copy files on multiple hosts simultaneously.

$ fuel-pdcp -r controller -- zabbix.deb /tmp

This will copy Zabbix package to /tmp on all controllers. Also, you may copy several files:

$ fuel-pdcp -r controller -- zabbix.deb zabbix.conf /tmp

Important: destination is considered directory. So if you do following

$ fuel-pdcp -r controller -- zabbix.deb /tmp/zabbix.deb

Then new directory /tmp/zabbix.deb/ will be created and you file gonna be copied in /tmp/zabbix.deb/zabbix.deb. This is intentional because to avoid ambiguaty on copying several files into one place. Please remember about that.