On 5/28/2022 10:19 PM, Kaushal Shriyan wrote:
I am running Apache Solr 8.11.1 on CentOS Linux release 7.9.2009 (Core)
using zookeeper to make it highly available. I have a couple of questions.
1. Is there a performance tuning to be carried out as I am using the
default settings?
Without knowing a lot of details about your setup that you may not want
to divulge, we can't make any recommendations about tuning. Any general
recommendations we might make could be completely wrong for your
particular setup.
A completely minimal redundant SolrCloud install consists of three
separate physical machines. Two of them would run zookeeper and solr as
separate processes, the third would just run zookeeper. If the third
machine only runs zookeeper, it would not need to have as much
horsepower as the other two. If you have the budget, maximum
performance will be achieved by NOT sharing nodes between solr and
zoookeeper. If you split Solr and ZK completely, you would need five
separate machines for a minimal install.
2. I am using 4 CPUs and 16 GB RAM for the 3 Apache Solr nodes. Is there
a way to benchmark or perform load tests on the Apache Solr Cluster?
Nothing built in or included in the Solr download. You can collect
queries and then write something or find something to send those queries
to Solr quickly. The solr.log file will by default contain every query
that Solr receives. If you harvest from there, be sure you know how to
figure out which queries are the actual query made by the user and which
are queries that SorlCloud sends from one node to another in order to
execute the query sent by the user.
Thanks,
Shawn
