Hi everyone,
as mentioned in my previous mail, we were going to re-import the osm
planet database for mapnik rendering, after it stopped updating in
September. This is now complete and the database is up-to-date again
with a lag of typically around 5 minutes behind the main osm database (
http://munin.toolserver.org/OSM/ptolemy/replication_delay2.html ).
At the same time there was a switch to postgresql 9.1, but all tools and
renderings should be working fine with the new database.
If anybody notices anything that doesn't work as intended, please let us
know to see if it can be fixed.
Not everything is working perfectly yet, but hopefully at least no worse
than before. The things currently not working correctly yet are:
1) Low-zoom rendering tiles. These are still very slow to render and
many of the low-zoom tiles still time out, even though I have set the
timeout up to 25 minutes. Not only does this clog up one of 3 rendering
slots for 25 minutes, it also results in that the tile isn't rendered
fresh and still considered "dirty", so the system puts those tiles into
the queue for rendering again immediately, clogging up the queue once
more. The current status of the queues can be seen either at
http://toolserver.org/~mazder/tirex-status/?short=1&extended=0&refr…
or
http://munin.toolserver.org/OSM/ptolemy/tirex_status_queued_requests.html
2) Tile expiry. Although the db is updated every 5 minutes, tiles are
not currently expired. For one, this is because there are some bugs in
the software (osm2pgsql) used to identify the tiles to expire. But also,
it is not yet clear how performance will be on trying to expire 250+
styles. We'll need to see what is feasible, but I suspect it will be
something like expiring the 10 or so most used styles with the db update
and then periodically expire the other styles with however much capacity
is still available.
We'll have to see how stable everything is and then move on from there.
Kai