www-infrastructure-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Joe Schaefer <joe_schae...@yahoo.com>
Subject Re: Best Practices so far?
Date Sat, 03 May 2008 00:00:33 GMT

--- Grzegorz Kossakowski <gkossakowski@apache.org>

> Justin Erenkrantz pisze:
> > The Fisheye guys know to contact us before
> importing history as they
> > know that they have routinely brought the machine
> down.  AFAICT, Ohloh
> > hasn't worked off SVN but rather off releases -
> and, to the best of my
> > knowledge, they haven't tripped any of our DoS
> measures...yet.
> I've asked at @infra mailing list every time I was
> going to try git-svn cloning of Cocoon 
> repository. The exception is the first attempt that
> I was sure will fail at the early beginning so 
> no load would be put on svn server.
> Ohloh provides[1] commits listing at regular basis
> so it must gather this information from our svn 
> server, right?

It does, and they are partly responsible for the
genesis of mod_dontdothat.  IIRC Ohloh was checking
out / from anonymizing proxies.  Without
we were powerless to stop them.

> > I only care as much as you don't place an
> unreasonable drain on our
> > public services.  If it takes 100x more compute
> and network power to
> > service git-svn than a normal svn client, then our
> capacity is blown
> > out of the water.
> Any client that pulls *whole* history of project is
> going to put some load on svn server. It's obvious.
> However, I'm going to do this only *once* in my life
> so we are really discussing something 
> exceptional that could be negotiated to be performed
> where the load is lower (e.g. at night). What's 
> more funny, I almost done this already using
> svn.eu.apache.org but something has got corrupted at
> the very end of the whole process.

You probably hit the bug we've been hoping will get
fixed soon, since it corrupts the repo.

> No disaster
> happened to svn.eu.apache.org that I'm aware of.

Right.  Too many svn log hits tho, which can be
mitigated by the flag we've been discussing.

> Tell me the limit of requests to svn server I can
> perform per hour and I'll tweak my toys. It will 
> take maybe two weeks to finish because of your limit
> but at least I won't distinguish from the crowd.

I wouldn't worry about tweaking it, to be honest.
I think there's a good chance we will be able to
support git-svn from committers, but we're a little
unsure about how and where to draw the lines.

Be a better friend, newshound, and 
know-it-all with Yahoo! Mobile.  Try it now.  http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ

View raw message