[Buildbot-devel] Few buildbot improvements

Dustin J. Mitchell dustin at zmanda.com
Sat Apr 10 20:24:04 UTC 2010

It sounds like you've done a lot of research, and have some good plans
laid out!  Thanks, in advance!

I'll chime in here, but others should also feel welcome to voice their opinions.

On Sat, Apr 10, 2010 at 12:56 PM, Tomas Čerkasas
<tomas.cerkasas at gmail.com> wrote:
> a) Fix ETA calculations. The current text-output based estimation doesn't
> quite work as expected: I guess "ETA 26s" switching to "ETA 4s" in 1s for a
> primitive "sleep 30" ShellCommand step deserves being called working "not
> quite as expected".

ETA is not an itch I've been inclined to scratch, so improvements
sound great so long as that do not overly burden those who are not
interested in it.  In my use of Buildbot, limited CPU resources mean
that the vast (20 to 1) majority of the time spent waiting for a
builder is in waiting for a slot on a VM to open up to execute it.

> b) Add some UI improvements: 1) Build's progressbar 2) Build's per-step time
> chart. Without diving into details of why/what/how the progressbar/charts
> could be implemented, first of all, I'd like to ask what's the project's
> attitude towards embedding any JavaScript library(jquery-ui, raphaeljs.com,
> or any other...) into the front-end? I know that's an important
> ethical-philosophical issue to choose between usability/sleek look and
> purism, but as some of the python's world
> (namely http://djangoadvent.com/1.2/jquery-admin/) already sold their souls
> for usability/sleek look, maybe that's not that evil thing to do after all?
> :)

The UI additions sound great!  Marcus was the prime force behind the
web frontend at this point, so I will defer to his opinion on adding
JS.  I have no objections.

I do want to be careful that these additional UI features do not
drastically increase the load on the buildmaster.  All web rendering
is essentially done synchronously, so lots of HTTP requests can starve
the buildmaster's attention to slaves.  Chris's work on slave-side log
buffering reduces the impact of this problem somewhat, but not


Open Source Storage Engineer

More information about the devel mailing list