8 minutes to fetch a tile? about vector-datasource HOT 6 CLOSED

8 minutes is unexpectedly high. Usually, the slowest part is the database query. I'd recommend logging slow queries in postgresql to see if anything comes up. If that is indeed the slow part, run an explain analyze on it; maybe it's missing an index.

from vector-datasource.

Ok cool, I'll try to gather some log data & figure out what's taking so long. I am running this on a pretty fast machine with around 24 cores. What sort of time should I expect for rendering a single tile?

from vector-datasource.

The slowest tiles should be on the order of seconds. When I run your curl example, it kicks back right away for me.

curl -s http://localhost:8080/buildings/16/19293/24641.mvt > /dev/null  0.01s user 0.00s system 4% cpu 0.183 total

Btw, I do have data in that tile, but I've loaded a nyc extract.

from vector-datasource.

Awesome, I'd like to get there!

from vector-datasource.

Okay, so it turns out it was a connection issue. I threw in a bunch of prints & found it was hanging on a connect call for this query: https://github.com/mapzen/TileStache/blob/integration-1/TileStache/Goodies/VecTiles/server.py#208-209.

So then diggin in a little, I found this: host – database host address (defaults to UNIX socket if not provided) - from psycopg2.connect documentation.

mapzen/vector-datasource/tilestache.cfg sets the host to localhost on all the dbinfo sections & causes super slow connections somehow, atleast with my current postgresql config. Not sure. But after removing the host & port entries in my config, it's lightning fast.

time curl http://localhost:5000/buildings/16/0/6.mvt
real 0m0.269s

from vector-datasource.

Glad that you were able to resolve it. Thanks for the update.

from vector-datasource.