Skip to content


Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Scalable cluster administration Python framework — Manage node sets, node groups and execute commands on cluster nodes in parallel.

Task: allow enabling/disabling MsgTree buffering at runtime

For now, it was only possible to disable task's internal output
gathering on stdout (or stderr) with task.set_default() when task was
not running. For example, before the first run:
    task.set_default("stdout_msgtree", False)

This patch allows programs to call task.set_default() at any time, to
disable or enable again output gathering. Note that by default,
gathering is enabled on stdout and also on stderr when it is separated
from stdout. Disabling internal output gathering when it's not useful
can save memory and CPU time.

Closes #181

Change-Id: I80d9d85eb34d24712b748a2115a77877b84a8930
latest commit c0dffa1176
thiell thiell authored


 ClusterShell 1.6 Python Library and Tools

ClusterShell is an event-driven open source Python library, designed to run
local or distant commands in parallel on server farms or on large Linux
clusters. It will take care of common issues encountered on HPC clusters, such
as operating on groups of nodes, running distributed commands using optimized
execution algorithms, as well as gathering results and merging identical
outputs, or retrieving return codes. ClusterShell takes advantage of existing
remote shell facilities already installed on your systems, like SSH.

ClusterShell's primary goal is to improve the administration of high-
performance clusters by providing a lightweight but scalable Python API for
developers. It also provides clush, clubak and nodeset, three convenient
command-line tools that allow traditional shell scripts to benefit from some
of the library features.

 Requirements (v1.6)

 * GNU/Linux, *BSD, Mac OS X, etc.

 * OpenSSH (ssh/scp)

 * Python 2.x (x >= 4)


ClusterShell is distributed under the CeCILL-C license, a French transposition
of the GNU LGPL, and is fully LGPL-compatible (see Licence_CeCILL-C_V1-en.txt).


When possible, please use the RPM/deb package distribution:

Otherwise in the source directory, use:

    # python install
    # cp -r conf /etc/clustershell

For installation on Mac OS X, please see:

 Test Suite

Regression testing scripts are available in the 'tests' directory:

    $ cd tests
    $ nosetests -sv <>
    $ nosetests -sv --all-modules

You have to allow 'ssh localhost' with no warning for "remote" tests to run.


Local API documentation is available, just type:

    $ pydoc ClusterShell

Online API documentation (epydoc) is available here:

 ClusterShell interactively

Python 2.7.1 (r271:86832, Apr 12 2011, 16:15:16)
[GCC 4.6.0 20110331 (Red Hat 4.6.0-2)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> from ClusterShell.Task import task_self
>>> from ClusterShell.NodeSet import NodeSet
>>> task = task_self()
>>>"/bin/uname -r", nodes="linux[4-6,32-39]")
<ClusterShell.Worker.Ssh.WorkerSsh object at 0x20a5e90>
>>> for buf, key in task.iter_buffers():
...     print NodeSet.fromlist(key), buf

linux[4-6] 2.6.32-71.el6.x86_64

 ClusterShell Tools

Powerful tools are provided: clush, nodeset and clubak.

* clush is a friendly and full-featured parallel shell (see: man clush).
  If in doubt, just check if your other parallel tools can do things like:
  # tar -czf - dir | clush -w node[10-44] tar -C /tmp -xzvf -

* nodeset is used to deal with your cluster nodeset, it can be bound to
  external groups (see: man nodeset and man groups.conf).

* clubak is a tool used to format output from clush/pdsh-like output
  (already included in clush with -b), see man clubak.


Main web site:

Github source respository:

Github Wiki:

Github Issue tracking system: project page:

Python Package Index (PyPI) link:

ClusterShell was born along with Shine, a scalable Lustre FS admin tool:


Stephane Thiell           <>
Aurelien Degremont     <>
Henri Doreau                 <>

CEA/DAM 2010, 2011, 2012 -
Something went wrong with that request. Please try again.