Hi,
Our es node is experiencing a search performance problem. A single search query to es (with no other load) could easily time out (30s) and the CPU can go > 100%. Is it because of the hardware problem? Here are some infos.
ES setup:
ES version: 5.2.2
Nodes: 1 node
Data size: ~8GB, ~800,000 docs
RAM: 8GB
CPU: 2 core intel
Hard drive: non-SSD
ES heap: ~3.5GB
The search query is with highlighting. It's searching against an index with ~35,000 docs.
If you highlight many results and looking to highlight multiple fields then the default highlighter is known to be expensive. In an effort to re-create the matching and scoring logic that went into producing each hit it does this pseudo code:
for each doc to be highlighted:
for each field that needs highlighting:
index field value into a temporary, in-memory index
run query against in-memory index
throw away in-memory index
You can see some of this in-memory indexing going on in your hot threads.
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.