[OSM-dev] tirex timeout - could not send data to client: Broken pipe

[Frederik Ramm]

Hi,

On 11/28/11 11:15, Petr Voldán wrote:
> We created own OSM server – unfortunately there is problem when we
> render metatiles in levels 6-10. Roughly one half of metatiles is not
> rendered due to tirex time-out. The second half and also tiles in level
> 11-18 are rendered correctly. We tried increase tirex timeout parameter
> (master_rendering_timeout=60), but it didn’t help. After that we found
> out problem in postgresql log:
>
> 2011-11-28 10:56:56 CET LOG: unexpected EOF on client connection
> 2011-11-28 10:56:56 CET LOG: could not send data to client: Broken pipe
> 2011-11-28 10:56:56 CET STATEMENT: SELECT AsBinary("way") AS
> geom,"landuse","natural","waterway" from ....

Do you have messages like "sending HUP to worker 'mapnik' with pid xxx 
(due to timeout)" in your /var/log/daemon.log? If yes, then the mapnik 
backend has timed out, probably because you forgot to set 
backend_manager_alive_timeout to a similarly high value (55 would be 
good in your case).

Also, try and make sure that when you pre-render stuff on critical zoom 
levels, you only have one rendering thread (i.e. enqueue your requests 
with a suitable priority like 99).

Bye
Frederik