[OSM-dev] Report on wiki.osm

Nick Hill nick at nickhill.co.uk
Sat Jan 13 19:25:24 GMT 2007



Frederik Ramm wrote:
> Hi,
> 
>> I don't see any reason to consider
>> occasional pages not being delivered as a fault
> 
> I'd say so too; but what you achieve by defensively configuring your
> Apache is that the occasional _request_ is not served (as opposed to
> the occasional _page_). For a page with many images (e.g. the "Map
> Features" article), this may make it next to impossible to load it
> with all images intact, prompting numerous reload attempts by the user
> hoping in vain to see the page without any broken images.

That's a good point, but once the web browser has achieved an http connection to 
an apache instance, it will in-line http requests over the already established 
TCP connection to that one server instance. The apache instance is likely to 
respond to all requests, unless the client delays more than 3 seconds between 
getting content elements. If we are regularly seeing pages with broken images, 
then that is a fault.

> 
>> Given that all visitors tot he domain www.openstreetmap.org are
>> redirected to the wiki front page, it may be an idea to take
>> snapshots, perhaps every half hour, of the wiki front page as the
>> front page for openstreetmap. This will reduce wiki load.
> 
> Sounds like a good idea. - I am not familiar with optimizing wikis,
> but on a different machine suffering from peak loads I once added a
> "%D" specification to the Apache LogFormat directive to make Apache
> log the number of milliseconds taken to serve each request:
> 
> LogFormat "%h %l %u %t \"%r\" %>s %b \"%{Referer}i\" \"%{User-Agent}i\" %D" combined-with-time

Interesting idea.




More information about the dev mailing list