I think the experimental flag can be removed from the documentation page.
It has been out in the wild for quite some time now. We keep on improving
the has_parent and the parent/child support in general. For example
recently in 0.90.1 we reduced the memory footprint of parent/child
significantly.
Besides the memory considerations, I would also recommend the use the
warmer api in combinations with parent/child. The allows the loading of the
id_cache to happen in a controlled manner as part of the refresh. If you
don't use warmers then all the initial first searches try the load the
values for the id_cache, but only one search request wins, so a lot
of unnecessary work is done as part of the first searches.
If you're planning on using parent/child on a index with a lot of documents
(like 100M docs), you might need to increase the number of primary shards
and have sufficient nodes for those shards in order to make searches the
parent/child queries execute within an acceptable time range.
Thanks Martijn. Good to know it's mature. That was my impression as well.
As for warmers, we are already using them because we are heavy users of
top_children, so the parent ids are already loaded up in RAM. I don't know
if has_parent will use the same cache but I think we have enough spare
memory to accomodate it as well. We haven't upgraded to 0.90.x yet but I
think that's in the works in the next month or so.
On Wednesday, June 26, 2013 5:50:12 AM UTC-4, Martijn v Groningen wrote:
I think the experimental flag can be removed from the documentation page.
It has been out in the wild for quite some time now. We keep on improving
the has_parent and the parent/child support in general. For example
recently in 0.90.1 we reduced the memory footprint of parent/child
significantly.
Besides the memory considerations, I would also recommend the use the
warmer api in combinations with parent/child. The allows the loading of the
id_cache to happen in a controlled manner as part of the refresh. If you
don't use warmers then all the initial first searches try the load the
values for the id_cache, but only one search request wins, so a lot
of unnecessary work is done as part of the first searches.
If you're planning on using parent/child on a index with a lot of
documents (like 100M docs), you might need to increase the number of
primary shards and have sufficient nodes for those shards in order to make
searches the parent/child queries execute within an acceptable time range.
On 25 June 2013 21:22, George Stathis <gsta...@gmail.com <javascript:>>wrote:
--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearc...@googlegroups.com <javascript:>.
For more options, visit https://groups.google.com/groups/opt_out.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.