Subject: Re: [boost] SVN either down or slowing to a crawl?
From: Vladimir Prus (vladimir_at_[hidden])
Date: 2010-08-06 07:58:31
Daniel James wrote:
>> From: DongInn Kim <dikim_at_[hidden]>
>> My name is DongInn Kim, systems administrator in the OSL. I have been
>> looking on the web server for a couple of days on wowbagger because of
>> the very high loadaverage on it.
>> The high loadaverage seems to be from most pages related to the zip
>> files under http://www.boost.org/doc/ because whenever the page is
>> accessed, the zip file(s) are decompressed and compressed back and
>> then the page can show their contents.
> Quick update, I'm working on a solution to this. It's mostly done, the
> delay is now just down to us living in different time zones. But the
> server's still very slow so there could be another issue. Although,
> that could also be due to the zipped pages that are still being served
> on the beta site and for regression test results. I'm deliberately
> breaking the beta site's documentation to see if that will help. I've
> also changed a couple of links to go to a copy of boost on sourceforge
> - although this is only a temporary measure, and still leaves a lot of
> broken links.
There's another issue with the server, it seems. When I do SVN commit,
it never finishes, e.g.:
ghost_at_wind:~/Work/Boost/boost-svn/tools/build/v2$ svn ci
Waiting for Emacs...
Transmitting file data ..
stays this way for many minutes (at least 10). However, I have already
received commit email about this commit. So, it seems like something
prevents SVN server from telling the client that commit is actually
done. Killing the client, together with 'svn cleanup', together
with 'svn up' works around this. This started to happen 3 days ago,
and I'm using at least two different network routes. I have svn 1.6.5
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk