[distcc] Multiple people running distcc at once,
using one DISTCC_DIR
vtnpgh at yahoo.com
Tue Sep 21 16:14:57 GMT 2004
Jean, Jake, et al.,
I am doing host randomization -- according to the outline I put in my "My Big
Plans for Distcc" posting a few weeks ago. So, we have good randomization
happening. But, we also have a very heterogenous collection of compilation
servers -- from very fast, to, shall we say, not so fast. So, I want
compilations to use the fast machines -- when they are not busy already.
I've also extended distcc so that it takes as input another file that has the
build machines' load averages in it. This load averages file is updated by
some python daemons that I've written. It all works pretty well.
But, I still have this problem of the lock files...
--- Jean Delvare <khali at linux-fr.org> wrote:
> On Tue, 21 Sep 2004 08:49:29 -0700, Jake McGuire wrote
> > Also, if you aren't randomizing the host list, you really want to do
> > so, and (IMHO) it's worth increasing the number of allowable jobs on
> > the remote machines so that you'll be unlikely to have lock
> > contention, as with 70 engineers and a networked lock directory
> > that'll get ugly, fast.
> As I understand it, once the host list is randomized, having a single,
> networked lock directory is pointless, isn't it? Just put a job limit on each
> daemon, in case the random is not random enough, and it should work fine.
> I think I remember that host list randomization is a new feature though,
> probably not present in 2.16, which the original poster is using.
> Jean Delvare
> distcc mailing list http://distcc.samba.org/
> To unsubscribe or change options:
Do you Yahoo!?
Take Yahoo! Mail with you! Get it on your mobile phone.
More information about the distcc