On 1/26/2015 2:58 AM, Dave F. wrote:
I don't know when it was last reviewed, but does this error have bit
of a sensitive trigger? Has the server that runs the process been
upgraded so it can handle a greater number of requests? If so, could
the error's cut in point be relaxed?
The thresholds for each server have been adjusted multiple times and
will probably continue to be so. Even if the load cutoff is increased
there will be times when individual render requests are rejected for a
few days in a row - stylesheet updates being the main one. It is
considered more important to update the map rendering than to do a
custom render. Personally I don't have many problems generating a custom
render from osm.org, but I'm on a different timezone and keep different
hours, which makes it hard to compare. I also get directed to a
different server much of the time.
For the technical details, the render code relies on the system load
average. This is not the best metric, a better one would be one of the
renderd supplied queue sizes (e.g. don't do custom renders if anything
is in the dropped queue) or to integrate it with renderd, but no one has
stepped forwards to code this.
The reason why there's the restriction in the first place is that doing
a custom render consumes 100x to 1000x the resources of a normal tile
request.
_______________________________________________
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk
