I have a very small Solr setup. The index is 32MB and there are only 8
fields, most of which are ints. I run a cron job every hour to use
DataImportHandler to do a full reimport of a database which has 42,600 rows.
There is minimal traffic on the server. Maybe a few dozen queries a
minute. Usually way less than 1 per second. They look like this:
INFO: [] webapp=/solr path=/select
params={sort=add_date+desc&fl=content_id&start=0&q=*:*&wt=json&fq=tag_id:((10)+AND+(8))&rows=180}
hits=35937 status=0 QTime=0
INFO: [] webapp=/solr path=/select
params={sort=add_date+desc&fl=content_id&start=0&q=*:*&wt=json&fq=tag_id:((10)+AND+(791+9))&rows=72}
hits=1651 status=0 QTime=6
INFO: [] webapp=/solr path=/select
params={sort=add_date+desc&fl=content_id&start=0&q=*:*&wt=json&fq=tag_id:((2)+AND+(10)+AND+(20+24+16)+AND+(31+32+33+792+793))&rows=250}
hits=6 status=0 QTime=1
QTime looks good. That's milliseconds, right?
Despite this, solr's java process is constantly using 100% or more CPU.
While writing this email I've seen it jump from 53% to 91% to 154%. It's
up and down all over the place.
I'm worried what might happen if the traffic load actually shot up. This
doesn't seem healthy.
I'm using the Jetty config from the example directory. Solr 3.5.0 straight
from apache.org.
# java -version
java version "1.6.0_22"
OpenJDK Runtime Environment (IcedTea6 1.10.6)
(amazon-52.1.10.6.44.amzn1-x86_64)
OpenJDK 64-Bit Server VM (build 20.0-b11, mixed mode)
Amazon EC2 running Amazon's standard "Amazon Linux" distribution (basically
CentOS)
Any advice?
Thanks
-jsd-
