Hello lists,
Again, Mz is new experimental package that is a lightweight set of Linux
bash scripts that provide both a CLI and a Web interface for systems
management of z/VM and Linux. These scripts work well between LPARs and
CECs. Again, a disclaimer: if you are expecting quality, well-tested
bug-free code, then this package is not for you (you have been warned :))
If you're the experimental, "bleading-edge" type, you might want to try it
out on a sandbox system.
An RPM and a PDF are available at:
http://sourceforge.net/projects/managing-z/
Each source file is *not* checked into svn/git as may be customary with
sourceforge - too many things to worry about for now, but that will
hopefully be coming..
A little background:
(1) Two interns worked at IBM back in the summer of '08. They created a
great Web front-end to a z/VM and Linux management interface. The problem
was that the back-end data structure was simply not there. It became
apparent that the data must come first - build the foundation before the
facade.
(2) One day I needed a chart to show the CEC - LPAR - z/VM -
VirtualMachine - Linux hierarchy and started fiddling with boxes, lines
and arrows. Not being a 'chart-meister', I thought how can to quickly draw
a hierarchy and the Linux 'tree' command came to mind. So I created a
hierarchy of Linux directories, ran "tree" and dropped the output in the
chart and was about to delete the directory structure when an "aha moment"
occured: Why can't the Linux file system be a database for the system z
hierarchy of resources? (Neale Ferguson once called it something like "the
/proc of System z resources")
(3) Another influence was a comment Marcy Cortez made years ago - words
the the effect of "I don't want a systems management solution that gives
me islands of automation (one island per z/VM LPAR), each of which need
care and feeding".
(4) Many customers have asked for "no root SSH" support: If root is not
permitted to SSH among systems, how does the systems management tool work?
I could go on, but it is these and other "zCommunity"
input/contributions/revelations from these lists, user group meetings and
conferences that have driven the design of Mz.
So what *is* Mz? To start with, it is a Linux RPM. A Linux system with the
Mz RPM installed is considered to be an "Mz server". It requires extra
privileges. Other Linux systems under the same z/VM can be managed by this
server. These are considered "Mz clients", however, there is no special
client code needed. The term for all Linux systems managed by Mz is nodes
(clients + servers = nodes). The Secure Shell (SSH) is used for
communication among Mz servers and clients and 'passwordless' SSHing is
important. Except for sshd, Mz is effectively "agentless". Mz servers
store data about all systems in a hierarchy of Linux directories and
files. This enables fast access to data and leverages the rich set of
tools that manipulate Linux file systems. There are three main branches of
this "Mz tree":
(1) Systems - the hierarchy of the System z resources: CECs, LPARs, z/VM
systems, virtual machines and Linux systems
(2) Appliances - imges of systems that have been 'captured' and can be
'deployed' to other z/VM LPARs
(3) SSI clusters - for z/VM 6.2 and later, Single System Image clusters
and members
Multiple Mz servers can join together to form an Mz cluster. This enables
cross-CEC and cross-LPAR systems management through SSH and TCP/IP
(hopefully addressing Marcy's concern). All servers in the cluster
synchronize with each other. When there are multiple Mz servers in the Mz
cluster, each server?s system data is duplicated on each of the other
servers but each Mz server "owns" the data on its LPAR.
First, second and third level Linuxes are supported (though first and
third have been only lightly tested).
This project has been a group effort that's been percolating for years. It
includes another excellent open-source tool, smaclient, written by Leland
Lucius of Sine Nomine Associates. Other IBMers have contributed to the
code, especially Christian Paro doing the initial capture and deploy work,
and recently, Marian Gasparovic in the area of Web serving (and he has
implemented some excellent ideas).
While IBM OK'd the open-sourcing of this and the redistribution of
smaclient, It is not an official IBM project. It's an experiment in
utilizing input from the community and the proven success of the open
source model. Nobody has their names in the code. While it could help IBM
sales and services, it is really being made available for you the customer
- to make it easier to manage your IT systems, and to make your
organizations more successful and efficient.
"Mike MacIsaac" <mikemac at-sign us.ibm.com>
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
----------------------------------------------------------------------
For more information on Linux on System z, visit
http://wiki.linuxvm.org/
