Clear the memory in Elasticsearch

Hi, everyone.

I am currently trying to find out how memory works in elasticsearch.
I am using BigDesk to monitor the heap space of my nodes, and it seems to
me that the heap memory only start to be freed when it is almost all used
up. Is that normal?
It seems to cause OOM errors when making big queries .
When the node has just started, they work fine (a bit lenghty, but the
cache effect hasn't kicked in yet, so nothing special).
But when several other queries have been made, I start getting OOM errors
that ultimately lead to a restart of the cluster.
The heap memory does not seem to clear with enough efficiency to make room
for the new query. Does that seem logical? Is there a setting somewhere to
define when to clear the memory? Is there a way to force-clear the heap
memory?
The flush API doesn't seem to do anything in that regard.

Thanks for any insights on the subject.

--

you miss a lot of details: how much memory do you have for elasticsearch,
how many docs/index size/..., what queries are you running etc

to be freed when it is almost all used up. Is that normal?

yes, that is the java way

Peter.

On Monday, December 10, 2012 10:54:33 AM UTC+1, DH wrote:

Hi, everyone.

I am currently trying to find out how memory works in elasticsearch.
I am using BigDesk to monitor the heap space of my nodes, and it seems to
me that the heap memory only start to be freed when it is almost all used
up. Is that normal?
It seems to cause OOM errors when making big queries .
When the node has just started, they work fine (a bit lenghty, but the
cache effect hasn't kicked in yet, so nothing special).
But when several other queries have been made, I start getting OOM errors
that ultimately lead to a restart of the cluster.
The heap memory does not seem to clear with enough efficiency to make room
for the new query. Does that seem logical? Is there a setting somewhere to
define when to clear the memory? Is there a way to force-clear the heap
memory?
The flush API doesn't seem to do anything in that regard.

Thanks for any insights on the subject.

--

Hi,

On Monday, December 10, 2012 4:54:33 AM UTC-5, DH wrote:

Hi, everyone.

I am currently trying to find out how memory works in elasticsearch.
I am using BigDesk to monitor the heap space of my nodes, and it seems to
me that the heap memory only start to be freed when it is almost all used
up. Is that normal?

That's typical, but you can control it with a few knobs like -XX:MinHeapFreeRatio,
-XX:SurvivorRation, -XX:TargetSurvivorRatio, etc.

Have a look
at http://www.oracle.com/technetwork/java/javase/tech/vmoptions-jsp-140102.html

Otis

ELASTICSEARCH Performance Monitoring - http://sematext.com/spm/index.html
Search Analytics - http://sematext.com/search-analytics/index.html

It seems to cause OOM errors when making big queries .
When the node has just started, they work fine (a bit lenghty, but the
cache effect hasn't kicked in yet, so nothing special).
But when several other queries have been made, I start getting OOM errors
that ultimately lead to a restart of the cluster.
The heap memory does not seem to clear with enough efficiency to make room
for the new query. Does that seem logical? Is there a setting somewhere to
define when to clear the memory? Is there a way to force-clear the heap
memory?
The flush API doesn't seem to do anything in that regard.

Thanks for any insights on the subject.

--