Cluster builds on windows

[email protected] Tue, 14 Sep 2004 17:22:43 -0700
Newsgroups gmane.comp.gnu.cons.general
Message-ID <OFBD13F8CE.AB308160-ON88256F10.00017DED-88256F10.00026FDB@philips.com>
This is a multipart message in MIME format.
--===============0973998019==
Content-Type: multipart/alternative;
	boundary="=_alternative 00026E2788256F10_="

This is a multipart message in MIME format.
--=_alternative 00026E2788256F10_=
Content-Type: text/plain; charset="us-ascii"

Has anyone done cluster builds with cons? I think the roblem can be 
divided into a few
parts:

1. Splitting the tasks in parallel (the pcons diffs do this for UNIX on a 
single mcahine)
2. Running the tasks on different machines
3. Making sure the source code is available on the different machines.

What's the best approach for 3? Should the slave machines cache the source
or is a network file system fast enough? I would probably start with a 
windows share and
see how it works.

What is a good strategy to distribute tasks? There are two main cases:

1: a real cluster connected with gigE or faster interconnect doing one 
build.

2: A lot of desktop machines connected with corporate ethernet doing 
several builds.
A developer starts a a build (or a scheduled build starts) and the build 
recruits available machines.

Has anyone ported pcons to Windows? Copying the unix do_command to th 
windows
section doesn't work. ActiveState perl complains about Bizzare error in 
fork().

Jonas
--=_alternative 00026E2788256F10_=
Content-Type: text/html; charset="us-ascii"


<br><font size=2 face="sans-serif">Has anyone done cluster builds with cons? I think the roblem can be divided into a few</font>
<br><font size=2 face="sans-serif">parts:</font>
<br>
<br><font size=2 face="sans-serif">1. Splitting the tasks in parallel (the pcons diffs do this for UNIX on a single mcahine)</font>
<br><font size=2 face="sans-serif">2. Running the tasks on different machines</font>
<br><font size=2 face="sans-serif">3. Making sure the source code is available on the different machines.</font>
<br>
<br><font size=2 face="sans-serif">What's the best approach for 3? Should the slave machines cache the source</font>
<br><font size=2 face="sans-serif">or is a network file system fast enough? I would probably start with a windows share and</font>
<br><font size=2 face="sans-serif">see how it works.</font>
<br>
<br><font size=2 face="sans-serif">What is a good strategy to distribute tasks? There are two main cases:</font>
<br>
<br><font size=2 face="sans-serif">1: a real cluster connected with gigE or faster interconnect doing one build.</font>
<br>
<br><font size=2 face="sans-serif">2: A lot of desktop machines connected with corporate ethernet doing several builds.</font>
<br><font size=2 face="sans-serif">A developer starts a a build (or a scheduled build starts) and the build recruits available machines.</font>
<br>
<br><font size=2 face="sans-serif">Has anyone ported pcons to Windows? Copying the unix do_command to th windows</font>
<br><font size=2 face="sans-serif">section doesn't work. ActiveState perl complains about Bizzare error in fork().</font>
<br>
<br><font size=2 face="sans-serif">Jonas</font>
--=_alternative 00026E2788256F10_=--



--===============0973998019==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
[email protected]
http://lists.gnu.org/mailman/listinfo/cons-discuss
Cons URL: http://www.dsmit.com/cons/
--===============0973998019==--