Problem with geo_points search queries

Hello, we are bulk indexing some geo_points informations and thats is going
ok. We encountering problem with this query to elastic:

{ "query" : { "filtered" : { "filter" : { "geo_distance" : { "coords" : [
53.441,
14.545
],
"distance" : "2km",
"distance_type" : "arc"
} },
"query" : { "match_all" : { } }
} },
"size" : 1
}

We have that Error:


We are observing huge jump in heap memory (set to 1gb min/max). This is
our settings (we are using clean elastic configuration becouse we cant
change production configuration for testing):

transport.tcp.compress: true
http.max_content_length: 1000mb
jmx.create_connector: true
index.cache.field.type: soft

without index.cache.field.type we encounter the same problem.

Our _status:

Sorry, for my English.

Hi,

In order to execute the geo related filters, the values for them need to
be loaded to memory, you simply need to increase the memory allocated to
elasticsearch or start more nodes in the cluster.

On Wed, Apr 11, 2012 at 4:31 PM, luc3k2202 luc3k2202@gmail.com wrote:

Hello, we are bulk indexing some geo_points informations and thats is
going ok. We encountering problem with this query to elastic:

{ "query" : { "filtered" : { "filter" : { "geo_distance" : { "coords" : [
53.441,
14.545
],
"distance" : "2km",
"distance_type" : "arc"
} },
"query" : { "match_all" : { } }
} },
"size" : 1
}

We have that Error:
http://pastebin.com/raw.php?i=FeNbQMcx
We are observing huge jump in heap memory (set to 1gb min/max). This is
our settings (we are using clean elastic configuration becouse we cant
change production configuration for testing):

transport.tcp.compress: true
http.max_content_length: 1000mb
jmx.create_connector: true
index.cache.field.type: soft

without index.cache.field.type we encounter the same problem.

Our _status:
http://pastebin.com/raw.php?i=ewZdfdTN

Sorry, for my English.

I'm having the same issue, but simply adding nodes is not really a viable
option for us.

We are hitting this error on a node with around 4 million records (of which
only around 40% have a location) and 2GB of RAM, but we expect to have
around 400 Million records. We can't plausibly run 100 nodes for 400
Million records. I had wondered if indexing the lat_lon on the field would
help but we also have more than one geo_point on some of the docs (around
20%) and according to the docs indexing lat_lon does not work if there's
more than one value for the field.

Is there any other approach that will get us there?

On Friday, April 13, 2012 5:08:55 AM UTC-7, kimchy wrote:

Hi,

In order to execute the geo related filters, the values for them need
to be loaded to memory, you simply need to increase the memory allocated to
elasticsearch or start more nodes in the cluster.

On Wed, Apr 11, 2012 at 4:31 PM, luc3k2202 <luc3...@gmail.com<javascript:>

wrote:

Hello, we are bulk indexing some geo_points informations and thats is
going ok. We encountering problem with this query to elastic:

{ "query" : { "filtered" : { "filter" : { "geo_distance" : { "coords" : [
53.441,
14.545
],
"distance" : "2km",
"distance_type" : "arc"
} },
"query" : { "match_all" : { } }
} },
"size" : 1
}

We have that Error:
http://pastebin.com/raw.php?i=FeNbQMcx
We are observing huge jump in heap memory (set to 1gb min/max). This is
our settings (we are using clean elastic configuration becouse we cant
change production configuration for testing):

transport.tcp.compress: true
http.max_content_length: 1000mb
jmx.create_connector: true
index.cache.field.type: soft

without index.cache.field.type we encounter the same problem.

Our _status:
http://pastebin.com/raw.php?i=ewZdfdTN

Sorry, for my English.

--