Re: Strange system behavior

[Paul Goyette <paul%whooppee.com@localhost>]

On Sun, 26 Sep 2010, Steven Bellovin wrote:

> > Well, nice try, but no luck.  I dropped the memory speed to 1067MHz
> > and it "survived" for much longer than usual (got about 95% of the
> > way through the 'build.sh release' job) but still failed.  This time
> > it was gcc that got the seg-fault rather than mandoc.
> >
> > I'm going to get some sleep.  The cron job will kick off a couple
> > more attempts over night;  it will be interesting if the morning
> > reveals that it now fails consistently in this new spot.
>
> But it's a really good clue that that is indeed the problem area.

Pretty much every time I've made a change in the past, the exact
location of the failure has changed.  Even when I replaced the hard
drive, the failure point moved.  And early on in this saga, gcc had
previously been the victim, too.  So I'm not sure that the latest
change is significant.

In any case, the two overnight runs actually ran to completion, and
that is the first time in a long time that I've seen back-to-back
runs without a failure.

Hopefully this is not a long-term solution.  With the memory running
at "normal" 1333MHz, a complete build-and-test pass takes 2.5 hours.
But with the reduction to 1067MHz, it takes 3.75 hours.  (I don't
understand how a 25% reduction in speed results in a 50% increase in
elapsed time...)


-------------------------------------------------------------------------
| Paul Goyette     | PGP Key fingerprint:     | E-mail addresses:       |
| Customer Service | FA29 0E3B 35AF E8AE 6651 | paul at whooppee.com    |
| Network Engineer | 0786 F758 55DE 53BA 7731 | pgoyette at juniper.net |
| Kernel Developer |                          | pgoyette at netbsd.org  |
-------------------------------------------------------------------------