(Longish, Friday afternoon musings).
We are running into some interesting problems in
testing Linux across platforms.
It is the opposite of building penguin colonies or
rookeries or rafts or waddles.
The scenario is common enough. We have a database
product that goes through release and fix cycles. We
then have several products that depend on the database
product with their own release and fix cycles
dependent on the database product. When we send out
release 487 of the db, then all other products have to
support release 487. Nothing really new hear.
Linux has added a new dimension to this with the
multiplicity of distributions and releases and the
multiplicity of platforms that support it. We develop
it say on intel and it seems to work. Does it work
on all other platforms and release levels we day it
does?
For example, just considering enterprise releases, we
have RH 7.1, 7.2, RH AS 2, RH AS3, RH AS3 U1, SLES7,
SLES7 SP1, SLES8, SLES8 SP3, the uncommon thread being
different vendor configurations, different kernel
levels, and different maintenance levels, some or all
of which run on our beloved dinosaurs, power PC's,
intel, itanium, Sun, and HP platforms (just the ones
we support now).
So what does the test matrix look like? A subset might
look like, with 23 different hosts required to
complete the testing.
intel intel ppc ppc zseries zeries
31bit 64 31 64 31 64
RH7.1 X
RH7.2 X
RH AS2 X
RH AS3 X X X X X X
SLES7 X X
SLES8 X X X X X X
SLES8-3 X X X X X X
And, obviously, the developers, testers, and users
want a common environment; they don't want to change
anything as they move from host to host. They want to
repeat the same certification on each host.
The question, then becomes, what do I have to do to
set up this environment? Some thoughts come to mind:
1 - each group needs a common storage for their work.
NFS appears to be the best choice because it tracks
distributions better than say AFS, which may not work
with bleeding edge products.
2 - With a common storage, userids and groups
(uid/gid) need to be consistent amongst all platforms
to access the data.
3 - The build of each host has to be essentially the
same as to products installed and not installed. (a
good example of this requirement is the compatibility
rpm dragged along from earlier distributions)
4 - There is an install test phase, where the
installation process needs to be certfied and this
invariably requires root access and an isolated system
host.
5 - There is a function test phase, where all the
users could conceivably use the same host, but we
would need hosts at each level.
So, does anyone have any thoughts on how to approach
this problem? Any references you know of?
(I don't like the word "challenge", I prefer the old
mathematical connotation of "problem"). ;-)
=====
Jim Sibley
RHCT, Implementor of Linux on zSeries
"Computer are useless.They can only give answers." Pablo Picasso
__________________________________
Do you Yahoo!?
Yahoo! Search - Find what you re looking for faster
http://search.yahoo.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
