Overpass API developpement

Text archives Help


Re: [overpass] std::bad_alloc runtime error


Chronological Thread 
  • From: mmd <mmd.osm@gmail.com>
  • To: overpass@listes.openstreetmap.fr
  • Subject: Re: [overpass] std::bad_alloc runtime error
  • Date: Sat, 23 Apr 2016 11:09:25 +0200

Hi Roland,

Am 20.04.2016 um 19:09 schrieb Roland Olbricht:

>>
>> Maybe the current limit is still a bit low for those kind of queries?
>
> I don't think so. The syscall tells us that the query is using 3 GB of
> RAM. That's a clear inidication that this query triggers a bug.
>
> Investigating this just yielded the fix
> e9ddc5fe7bdfb3f3382449489e22c9337e3377f1.
>
> A desired side effect is that this query now completes much faster.
>

Great, thanks!

> In fact, this is not the whole story, but it helps anyway. The query
> statement should be smart enough to recognize that there are much to
> many elements tagged "building" to load them. But to fix this is a
> different story and would need more testing.

While we're at the topic, here's another report [1] from someone who
tried to query for some buildings in a very small bbox at zoom level 19.
It triggered a "runtime error: Query run out of memory using about 2048
MB of RAM."

way ["building"="yes"] ({{bbox}});
(._;>;);
out;

http://overpass-turbo.eu/s/fPs

So, again to my point, I believe 2 GB is a bit too restrictive and I
would still recommend to go for 4 GB for the time being. I see the hard
mem limit as a line of defense against really excessive memory
consumption. Once additional patches to reduce memory consumption are
in, we can think about reducing the memory limit further if that's still
needed.

br,
mmd


[1] http://forum.openstreetmap.org/viewtopic.php?id=54407



Archive powered by MHonArc 2.6.18.

Top of page