esperf - Elasticsearch Stress Testing Tool

esperf is a multi-thread program designed for measuring Elasticsearch cluster's search, aggregation and other request capacity. It reads the query DSL from the standard input and performs HTTP requests as the request body to the specified URL. It is also able to modify the query string with random numbers and random strings in each request.

Usage: esperf [-d dictionary_file] [-r max_recurrence] [-t num_threads] [-X method] [-u user:password] url Options:

• -d dictionary_file: Newline delimited strings dictionary file
• -h: Show this help
• -r max_recurrence: Number of recurrence HTTP requests per thread (default 10)
• -t num_threads: Number of threads to generate, not always a big number gives more pressure (default 5)
• -u user:password: Username and password for HTTP authentication
• -v: Verbose outputs for debugging purpose
• -X: HTTP method to perform

Output format

<ISO 8601 formatted timestamp> <number of successful requests/sec> <number of failures/sec>

Example usages

Simply retrieve the greeting message 50 times.

$ ./esperf "http://localhost:9200"

Perform a match_all query 1000x3 times on all shards from 3 threads.

$ echo '{"query": {"match_all": {}}}' | ./esperf -r 1000 -t 3 "http://localhost:9200/_search"

Perform range queries with randomly generated numbers.

$ echo '{"query": {"range": {"my_length": {"gte": $RNUM}}}}' |  ./esperf -r 1000 -t 3 "http://localhost:9200/_search"

Perform term queries with randomly selected strings from the dictionary.

$ echo '{"query": {"term": {"first_name": {"value": "$RDICT"}}}}' | ./esperf -r 1000 -t 3 -d ./names.txt "http://localhost:9200/_search?size=1"

Perform bulk insert requests.

$ ./esperf -X PUT -r 1000 -t 3 -d ./names.txt "http://localhost:9200/test-index/test-type/_bulk" < bulk.txt

Your may alo refer to ibcurl error codes for curl_easy_perform() related errors.

Downloads

• esperf-0.1.2-linux-x86_64.zip
• esperf-0.1.2-darwin-x86_64.zip

How to build

Platforms

The program is confirmed built and run on the following operating systems.

• Linux x86_64
• Darwin x86_64 (OS X)

Dependencies

• cmake > 2.8
• gcc-g++ > 4.8
• libcurl-devel > 7.40

Make

Simply run the below commands to generate the executable on the target platform.

$ cmake ./CMakeList.txt
$ make

TODO:

• Provide more variations of random numbers and the geo type