[distcc] DMUCS 0.1 is available on sourceforge.net

[Sean D'Epagnier]

This message is mostly for victor, but feel free to read and comment.

I tested dmucs out. I'm not sure what's going on exactly, I don't know if
it's working right or not. I made a hosts-info file, and edited
dmucs_hosts_file.h and recompiled.  Maybe dmucs can have a command line
option to specify this file instead?

Then I ran dmucs on host a. I had to run loadavg -s localhost (otherwise it
would not find the server)  I think this should be the default action, or it
should make -s required.  I ran gethost, and it gave me back
127.0.0.1(using -D flag)  I compiled everything on host b (not
bothering to edit any
files) and ran loadavg -s other (other resolves to host a)  I then wrote a
simple program called roll that puts the cpu in an infinite loop (to consume
cpu) and ran it on host b.  Now dmucs and loadavg is running on host a.
loadavg and roll is running on host b.  I ran gethost on host b thinking I
should not get 127.0.0.1 but instead the hostname or ip of host a since it
is idle and host b isn't.  I got 127.0.0.1  so what did I do wrong?

I have a possible idea, it's not too serious, tell me what you think.  I
like the idea of having server (dmucs) and agent (loadavg) the same
program.  Now instead of a client-server model, you can build tree-like
structures.  This way you can have a dmucs server running for groups of
computers this feature might not be needed for most users, but it can be
useful for large networks.

in loadavg.cc:
Why define a sleep function (it's already there) and instead of sending the
load average at an interval of 10 seconds, the load average should update
much faster, but only when the rate of change is high.  If dmucs returns a
certain hostname to gethost, maybe it should poll that client in a second or
so as well.  If everything is idle, no need to say anything.  Getting load
average data from the uptime command is not a good idea imho.  I will look
into getting this data directly from the kernel, and not averaged data over
a period, but immediate data.

Is it true that if you have two 500mhz cpus in a computer and another
1000mhz single cpu the power factor should be 1 for both?  Or shoule the
1000mhz cpu be 2?  I think this should be made clear, and an example
hosts-info (with comments) be distributed.

I think this dmucs can do more.  It can be used to distribute environment
variables.  Instead of specifying distcc x86_64-pc-linux-gnu-gcc  I could
have dmucs set the correct env variables for distcc to do this.  Before this
gets too big, we should abstract the possibilites of dmucs (so it will be
made in such a way that it can be used with more than just distcc)

sean

On 1/12/06, Victor Norman <vtnpgh at yahoo.com> wrote:
>
> All,
>
> I'm excited to announce that DMUCS release 0.1 is available on
> sourceforge.net.  See http://dmucs.sourceforge.net for lots of information
> about how it works, etc.
> Download the tarball from http://sourceforge.net/projects/dmucs.
>
> I've very interested in getting more feedback on this project.  Any
> developers that wish to contribute are very welcome too.
>
> Thanks.
>
> Vic
>
> ------------------------------
> Yahoo! Photos
> Ring in the New Year with Photo Calendars<http://us.rd.yahoo.com/mail_us/taglines/photos/*http://pa.yahoo.com/*http://us.rd.yahoo.com/mail_us/taglines/photos/evt=38087/*http://pg.photos.yahoo.com/ph//page?.file=calendar_splash.html&.dir=>.
> Add photos, events, holidays, whatever.
>
>
> __
> distcc mailing list            http://distcc.samba.org/
> To unsubscribe or change options:
> https://lists.samba.org/mailman/listinfo/distcc
>
>
-------------- next part --------------
HTML attachment scrubbed and removed