chlorochrule/solrcloud-development-box

A Docker image for SolrCloud development

Note: This repository is for Lucene and Solr developers. DO NOT USE AS PRODUCTION.

Development environment for SolrCloud using Docker and Docker compose. This repository provides customized build for any lucene commit and solr commit.

How to use: Setup SolrCloud

prepare json data

Amazon Shopping Queries Dataset is able to be got from here.

This dataset is included as a submodule in this repository.

You can prepare json data by the following steps:

take the data from github, this step may take a long time
```
git submodule update --init --recursive
```

generate dataset

cd data
python3 create_shopping_data.py
cd ..

then you can check you json data in folder data/amazonshopping

Launch SolrCloud

Run:

docker-compose up -d --build

You can see the solr system by accessing localhost:8983 or localhost:8984.

The pull down menu "Collection Selector" is empty now.

Upload configsets and create collection

Configsets of amazon shopping data collection can be uploaded by:

# upload configsets
docker-compose exec solr1 /opt/solr/server/scripts/cloud-scripts/zkcli.sh -zkhost zoo:2181 -cmd upconfig -confdir /opt/solr/configsets/amazonshopping -confname amazonshopping
# create collection
curl 'localhost:8983/solr/admin/collections?action=CREATE&name=amazonshopping&numShards=2&replicationFactor=2&maxShardsPerNode=2&collection.configName=amazonshopping'

The file solr/configsets/amazonshopping is created base to the json file in data/amazonshopping/**.json,

The field defines every fields in json file, with the primary key named example_id.

Now, access to localohst:8983, you can find the amazonshopping included in the pull down menu "Collection Selector".

※ You can find Solr API here

Feed documents

The json file prepared before can be feed by curl command:

# Feed all documents
for docs in $(ls ./data/amazonshopping/*.json); do curl -X POST -H 'Content-Type: application/json' --data-binary @$docs 'http://localhost:8983/solr/amazonshopping/update?commit=true'; done

Access to http://localhost:8983/solr/#/amazonshopping/query, press Execute Query button and you can randomly get 10 json data.

How to use: Benchmark using JMeter

Create Solr query

mkdir -p jmeter/benchmarks
mkdir -p jmeter/benchmarks/benchmark-solr
cd data
python3 create_shopping_query.py > ../jmeter/benchmarks/queries.txt
cd ..

Run benchmark-solr script

docker-compose exec -u root jmeter benchmark-solr -c amazonshopping -z zoo:2181 -q /var/jmeter/benchmarks/queries.txt -t 233 -d 60 --extract-expression '$.response.numFound' --extract-expression '$.responseHeader.QTime' --clean

See report.txt

cat jmeter/benchmarks/benchmark-solr/*/report.txt

report.txt example

--- summary report ---
num requests: 6578
execution time: 59.671 [sec]
error (without timeout) count: 0
timeout count: 0
error (with timeout) rate: 0.0
throughput: 110.23780395837174 [rps]
           elapsed      bytes  sentBytes  Latency
min           4.00     679.00     169.00     4.00
50%ile       48.00   44198.00     186.00    47.00
75%ile      180.00   62605.75     186.00   177.00
95%ile      734.15  104636.80     204.00   728.15
99%ile     1162.92  161320.78     222.00  1157.69
99.9%ile   1557.27  272524.30     240.00  1557.27
99.99%ile  1894.05  321512.07     258.00  1872.00
max        1971.00  330710.00     258.00  1970.00
mean        159.69   50264.35     186.80   157.57
std         248.78   30548.91       9.68   247.28
--- end report ---

Java Flight Recorder during benchmarking

docker-compose exec -u solr solr1 start-jfr.sh 60  # 60sec
ls solr/mnt/solr1/benchmarks/solr.jfr

Graceful Shutdown

docker-compose exec -u solr solr1 solr stop -all
docker-compose exec -u solr solr2 solr stop -all
docker-compose down

Prometheus

Access: localhost:9090

Grafana

Accesslocalhost:3000
Create Data sources of Prometheus (URL is http://prometheus:9090)
Import dashboard

Customize for your development

Fix below environment variables:

ARG LUCENE_REPOSITORY="https://github.com/chlorochrule/lucene"  # Lucene repository you want to build
ARG LUCENE_CHECKOUT="main"  # Lucene branch, commit or tag

ARG SOLR_REPOSITORY="https://github.com/chlorochrule/solr"  # Solr repository you want to build
ARG SOLR_CHECKOUT="SNAPSHOT"  # Solr branch, commit or tag

How to refer Lucene SNAPSHOT from Solr

When referencing a particular Lucene branch, for example main, from Solr, Gradle build settings of Solr must be fix.

Fix gradle/globals.gradle.

@@ -24,6 +24,7 @@ allprojects {

   // Repositories to fetch dependencies from.
   repositories {
+    mavenLocal()
     mavenCentral()
     maven {
       name "LucenePrerelease${lucenePrereleaseBuild}"

Fix version.props.

@@ -87,7 +87,7 @@ org.apache.httpcomponents:httpmime=4.5.10
 org.apache.james:apache-mime4j*=0.8.3
 org.apache.kerby:*=1.0.1
 org.apache.logging.log4j:*=2.13.2
-org.apache.lucene:*=9.1.0
+org.apache.lucene:*=9.3.0-SNAPSHOT
 org.apache.opennlp:opennlp-tools=1.9.1
 org.apache.pdfbox:*=2.0.17
 org.apache.pdfbox:jempbox=1.8.16

Update dependencies.

./gradlew --write-locks

Check build success.

./gradlew assemble

License

Apache-2.0 License

Author

Naoto Minami

chlorochrule / solrcloud-development-box