Massive "bzr update"

Massive "bzr update"

[Eli Zaretskii]

I just resync'ed with the repository, and bzr again pulled almost
200MB through the wire.  It took 21 minutes.  I watched the CPU meter:
it stayed below 10% almost all the time, with a few peaks at 30%.  The
memory footprint went steadily up to 300MB, then stayed there for most
of the time; that is not too bad on my 1.5GB box.  So the only
bottleneck could be network throughput (pulling 200MB should take
around 15 minutes for my connection, which is a large portion of the
21 minutes it took).  But 3.5Mbps connection is not exactly a slow
one, so I don't expect weekly updates to be that slow.

The files it updated were just a few:

 M  doc/lispref/ChangeLog
 M  doc/lispref/control.texi
 M  doc/lispref/display.texi
 M  doc/lispref/os.texi
 M  doc/lispref/positions.texi
 M  doc/lispref/text.texi
 M  doc/lispref/variables.texi
 M  doc/misc/ChangeLog
 M  doc/misc/calc.texi
 M  doc/misc/gnus.texi
 M  lisp/ChangeLog
 M  lisp/font-setting.el
 M  lisp/gnus/ChangeLog
 M  lisp/gnus/gnus-art.el
 M  lisp/gnus/gnus-sum.el
 M  lisp/gnus/gnus-util.el
 M  lisp/gnus/message.el
 M  lisp/international/fontset.el
 M  lisp/language/indian.el
 M  lisp/progmodes/python.el
 M  lisp/vc-bzr.el
 M  lisp/vc-dir.el
 M  src/ChangeLog
 M  src/dbusbind.c
 M  src/font.c
 M  src/font.h
 M  src/fontset.c
 M  src/gtkutil.c
 M  src/keyboard.c
 M  src/nsfns.m
 M  src/xftfont.c
 M  src/xterm.c

My previous "bzr up" was 6 days ago.  I'm used to update my tree once
a week, because I only have time to work on Emacs on weekends.  Is
this massive and slow update something I should look forward to, until
Savannah admins enable bzr+ssh?  Or is this again something
``exceptional'', like the problem we had last week?

For comparison, a "cvs up" would take about 1/10 of that time, even
though my CVS tree was checked-out with -kb (to avoid EOL
conversions), so it would send entire files downstream instead just
the patches.

Re: Massive "bzr update"

[Óscar Fuentes]

Eli Zaretskii <eliz@gnu.org> writes:

> I just resync'ed with the repository, and bzr again pulled almost
> 200MB through the wire.

Does that "again" means that you experimented a previous "update" that
required 170 MB?

Some days ago there was an issue with an update that took about that
same size. I just updated and it was as fast (or slow) as usal (30
seconds).

-- 
Óscar

Re: Massive "bzr update"

[Eli Zaretskii]

> From: Óscar_Fuentes <ofv@wanadoo.es>
> Date: Fri, 08 Jan 2010 14:59:39 +0100
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > I just resync'ed with the repository, and bzr again pulled almost
> > 200MB through the wire.
> 
> Does that "again" means that you experimented a previous "update" that
> required 170 MB?

Yes.  I reported that one as well.

> I just updated and it was as fast (or slow) as usal (30 seconds).

When was your previous update?  Mine was a week ago, as I told.

Re: Massive "bzr update"

[Óscar Fuentes]

Eli Zaretskii <eliz@gnu.org> writes:

>> > I just resync'ed with the repository, and bzr again pulled almost
>> > 200MB through the wire.
>> 
>> Does that "again" means that you experimented a previous "update" that
>> required 170 MB?
>
> Yes.  I reported that one as well.

Ouch.

>> I just updated and it was as fast (or slow) as usal (30 seconds).
>
> When was your previous update?  Mine was a week ago, as I told.

Dunno. I suspect that the problem is not about the raw amount of pending
changes, but about some automatic reorganization of the remote branch
data that forces the remote bzr client to read a large amount of data.

-- 
Óscar

Re: Massive "bzr update"

[Stefan Monnier]

> Is this massive and slow update something I should look forward to,
> until Savannah admins enable bzr+ssh?  Or is this again something
> ``exceptional'', like the problem we had last week?

IIUC it's "something I should look forward to,
until Savannah admins enable bzr+ssh".


        Stefan

Re: Massive "bzr update"

[Eli Zaretskii]

> From: Óscar_Fuentes <ofv@wanadoo.es>
> Date: Fri, 08 Jan 2010 18:48:48 +0100
> 
> >> I just updated and it was as fast (or slow) as usal (30 seconds).
> >
> > When was your previous update?  Mine was a week ago, as I told.
> 
> Dunno. I suspect that the problem is not about the raw amount of pending
> changes, but about some automatic reorganization of the remote branch
> data that forces the remote bzr client to read a large amount of data.

Well, let's see what happens next week (if bzr+ssh isn't enabled by
then).

Thanks for the feedback.