[Piuparts-devel] Thoughts on masterd
Holger Levsen
holger at layer-acht.org
Sun Dec 4 10:55:29 UTC 2011
Hi Dave,
On Samstag, 3. Dezember 2011, Dave Steele wrote:
> On Sun, Nov 27, 2011 at 4:25 PM, Dave Steele <dsteele at gmail.com> wrote:
> > ...couple of problems with piuparts-master:
> > I've been thinking ...
> > ...idea:
> > ...humbly propose...
> > Thoughts? ...
>
> It seems that I don't have a clear sense of what the consensus goal is
> on this topic, I'll concentrate elsewhere.
of course you're free to come to this conclusion, but let me assure you that
I'm all for making piuparts-master faster. And maybe your proposal is the way
to go! I just don't know...
The problem is, your plans are quite intrusive and we have a reliably working
master atm. So if you want me to replace that code with something else, you
need to show me the code, not just a mere proposal.
As I see it, improving startup speed of master would gain/free 30-60 cpu
minutes on piatti per day, at the maximum. While the bottleneck really is disk
I/O, so this would gain us little, while it introduces instability & work.
(And you might have noticed that sadly my time for piuparts atm is rather
limited too.)
You might want to check
http://munin.debian.org/debian.org/piatti.debian.org/index.html
All this said, having a masterd running all the time sounds like a good idea
(as long as its not eating too much memory, as memory is used to cache disk-
data thus saving I/O.)
I'm unsure whether to run a second slave on piatti (it has a dualcore cpu),
but then the real/worse bottleneck is processing the generated logs: there are
almost 500 detected but unreported bugs in sid and 100 in wheezy! :( (*)
So sure, we could do better at generating them faster and I'm happy to include
such patches, but I want do concentrate on filing bugs (and detecting new
types of issues).
(And piuparts-report should not take a quarter of the day / must run faster so
we can add more sections + suites. But we just cut that down from 6h to 4h, so
I'm inclined to add squeeze2wheezy tests soon.)
But as said: give me patches which are proven to work, and I'll be thankful
and gladly merge them! :-)
cheers,
Holger
(*) sure, some people directly look at the piuparts results. some even run
piuparts. but to get most bugs fixed, bugs need to be filed first.
More information about the Piuparts-devel
mailing list