Branch | Build Status |
---|---|
master |
This repository has common artifacts to be used for Performance Tests.
In components, there are several Java projects and each project builds an executable JAR file.
The distribution directory has the scripts and the Maven project to build the final distribution package including all scripts and components to be used for performance tests.
The package (performance-common-distribution-${version}.tar.gz) built by the distribution maven module can be used as a dependency to write performance testing scripts.
This package only provides helper scripts and applications. You must extend the functionality of these scripts to run performance tests.
It's recommended to include the contents of this package with any scripts written to extend the functionality.
Following is the tree view of the contents inside distribution package.
ubuntu@server:~$ tree --charset=ascii | sed -e 's/[0-9]\.[0-9]\.[0-9].*\.jar/${version}.jar/g'
.
|-- cloudformation
| |-- cloudformation-common.sh
| |-- create-template.py
| |-- download-logs.sh
| |-- get-wum-updated-wso2-product.sh
| |-- python-requirements.txt
| `-- templates
| `-- common_perf_test_cfn.yaml
|-- common
| `-- common.sh
|-- java
| `-- install-java.sh
|-- jmeter
| |-- create-summary-csv.sh
| |-- create-summary-markdown.py
| |-- csv-to-markdown-converter.py
| |-- install-jmeter.sh
| |-- jmeter-server-start.sh
| |-- perf-test-common.sh
| `-- user.properties
|-- jtl-splitter
| |-- jtl-splitter-${version}.jar
| `-- jtl-splitter.sh
|-- netty-service
| |-- netty-http-echo-service-${version}.jar
| `-- netty-start.sh
|-- payloads
| |-- generate-payloads.sh
| `-- payload-generator-${version}.jar
|-- sar
| `-- install-sar.sh
`-- setup
|-- setup-common.sh
|-- setup-jmeter-client.sh
|-- setup-jmeter.sh
`-- setup-netty.sh
10 directories, 26 files
Each directory has one or more executable scripts. All scripts support -h
(help) option.
Note: Most of these scripts will work only on Debian based systems like Ubuntu.
See following sections for more details.
The performance test scripts can create AWS resources to run standard server performance tests on AWS.
Standard server performance tests include a Back-end service and the parameters include "Application heap memory sizes", "Concurrent users", "Message sizes", and "Back-end Sleep Times".
The create-template.py
script can create AWS CloudFormation template based on a Jinja2
template. The common_perf_test_cfn.yaml
file
is a common Jinja2
template, which has necessary configurations to create a performance test environment on AWS.
You should extend cloudformation-common.sh
to run standard server performance tests. By extending, you can will get all the functionality
of creating the AWS stack, run the tests, delete stack and summarize results. You can even run several stacks to run performance tests parallely.
The cloudformation-common.sh
script is not directly executable, but it supports following options.
ubuntu@server:~$ ./cloudformation/cloudformation-common.sh -h
Usage:
./cloudformation/cloudformation-common.sh -f <performance_scripts_distribution> [-d <results_dir>] -k <key_file> -n <key_name>
-j <jmeter_distribution> -o <oracle_jdk_distribution> -g <gcviewer_jar_path>
-s <stack_name_prefix> -b <s3_bucket_name> -r <s3_bucket_region>
-J <jmeter_client_ec2_instance_type> -S <jmeter_server_ec2_instance_type>
-N <netty_ec2_instance_type>
[-t <number_of_stacks>] [-p <parallel_parameter_option>] [-w <minimum_stack_creation_wait_time>]
[-h] -- [run_performance_tests_options]
-f: Distribution containing the scripts to run performance tests.
-d: The results directory. Default value is a directory with current time. For example, results-20190124094422.
-k: Amazon EC2 Key File. Amazon EC2 Key Name must match with this file name.
-n: Amazon EC2 Key Name.
-j: Apache JMeter (tgz) distribution.
-o: Oracle JDK distribution.
-g: Path of GCViewer Jar file, which will be used to analyze GC logs.
-s: The Amazon CloudFormation Stack Name Prefix.
-b: Amazon S3 Bucket Name.
-r: Amazon S3 Bucket Region.
-J: Amazon EC2 Instance Type for JMeter Client.
-S: Amazon EC2 Instance Type for JMeter Server.
-N: Amazon EC2 Instance Type for Netty (Back-end) Service.
-t: Number of stacks to create. Default: 1.
-p: Parameter option of the test script, which will be used to run tests in parallel.
Default: u. Allowed option characters: ubsm.
-w: The minimum time to wait in minutes before polling for cloudformation stack's CREATE_COMPLETE status.
Default: 5.
-h: Display this help and exit.
When running the performance tests, you can use get-wum-updated-wso2-product.sh
script to get a WSO2 product with all WUM updates.
Use the install-java.sh
script to install Oracle Java Development Kit (JDK) on 64bit Linux.
The install-java.sh
script in this directory will not be useful when OpenJDK is used. It's recommended to use the default package
repositories to install OpenJDK.
Currently install-java.sh
script supports installing Oracle JDK 8.
You must download latest JDK 8.
This script can also install Java Cryptography Extension (JCE) Unlimited Strength Jurisdiction Policy files. You need to copy JCE Policy zip file to the same location as the downloaded JDK file (tar.gz)
The script needs to be run as root. The JDK will be extracted to /usr/lib/jvm
directory.
ubuntu@server:~$ ./java/install-java.sh -h
Usage:
./java/install-java.sh -f <java_dist> [-p <java_dir>] [-u <user>] [-h]
-f: The jdk tar.gz file.
-p: Java installation directory.
-u: Target user. Default: ubuntu.
-h: Display this help and exit.
The install-jmeter.sh
script will extract JMeter, install Plugin Manager and copy the user.properties
file.
The user.properties
file has recommended configurations for performance tests.
There is an option to download latest Apache JMeter.
ubuntu@server:~$ ./jmeter/install-jmeter.sh -h
Usage:
./jmeter/install-jmeter.sh -i <installation_dir> [-f <jmeter_dist>] [-d] [-p <jmeter_plugin_name>] [-h]
-i: Apache JMeter installation directory.
-f: Apache JMeter tgz distribution.
-d: Download Apache JMeter from web.
-p: The name of the JMeter Plugin to install. You can provide multiple names.
-h: Display this help and exit.
You shoud extend perf-test-common.sh
to run standard server performance tests. It supports testing with multiple concurrent
users, different message sizes, different Back-end service delays and different heap memory sizes of the target server.
The script also supports running remote (distributed) JMeter tests and it will also summarize the results for warmup and measurement periods.
A script can extend this perf-test-common.sh
script by simply sourcing the perf-test-common.sh
script.
For example:
# Execute common script
. $script_dir/perf-test-common.sh
Any script depending on this script must define test scenarios as follows:
declare -A test_scenario0=(
[name]="test_scenario_name1"
[display_name]="Test Scenario 1"
[description]="Description of Test Scenario 1"
[jmx]="test_scenario_name1.jmx"
[use_backend]=true
[skip]=false
)
declare -A test_scenario1=(
[name]="test_scenario_name2"
[display_name]="Test Scenario 2"
[description]="Description of Test Scenario 2"
[jmx]="test_scenario_name2.jmx"
[use_backend]=true
[skip]=false
)
Then define following functions in the script.
initialize
before_execute_test_scenario
after_execute_test_scenario
In above functions, following variables may be used
scenario_name
heap
users
msize
sleep_time
report_location
Use jmeter_params
array in before_execute_test_scenario
to provide JMeter parameters.
In before_execute_test_scenario JMETER_JVM_ARGS
variable can be set to provide
additional JVM arguments to JMeter.
Finally, execute test scenarios using the function test_scenarios
.
ubuntu@server:~$ ./jmeter/perf-test-common.sh -h
Usage:
./jmeter/perf-test-common.sh -m <heap_sizes> -u <concurrent_users> -b <message_sizes> -s <sleep_times> [-d <test_duration>] [-w <warmup_time>]
[-n <jmeter_servers>] [-j <jmeter_server_heap_size>] [-k <jmeter_client_heap_size>] [-l <netty_service_heap_size>]
[-i <include_scenario_name>] [-e <include_scenario_name>] [-t] [-p <estimated_processing_time_in_between_tests>] [-h]
-m: Application heap memory sizes. You can give multiple options to specify multiple heap memory sizes. Allowed suffixes: M, G.
-u: Concurrent Users to test. You can give multiple options to specify multiple users.
-b: Message sizes in bytes. You can give multiple options to specify multiple message sizes.
-s: Back-end Sleep Times in milliseconds. You can give multiple options to specify multiple sleep times.
-d: Test Duration in seconds. Default 900.
-w: Warm-up time in seconds. Default 300.
-n: Number of JMeter servers. If n=1, only client will be used. If n > 1, remote JMeter servers will be used. Default 1.
-j: Heap Size of JMeter Server. Allowed suffixes: M, G. Default 4G.
-k: Heap Size of JMeter Client. Allowed suffixes: M, G. Default 2G.
-l: Heap Size of Netty Service. Allowed suffixes: M, G. Default 4G.
-i: Scenario name to to be included. You can give multiple options to filter scenarios.
-e: Scenario name to to be excluded. You can give multiple options to filter scenarios.
-t: Estimate time without executing tests.
-p: Estimated processing time in between tests in seconds. Default 60.
-h: Display this help and exit.
Use create-summary-csv.sh
to create a summary CSV file.
ubuntu@server:~$ ./jmeter/create-summary-csv.sh -h
Usage:
./jmeter/create-summary-csv.sh -n <application_name> [-c <column_header_name>] [-r <regex>] [-x]
[-p <file_prefix>] [-g <gcviewer_jar_path>] [-d <results_dir>]
[-j <jmeter_servers>] [-k <application_instance_count>] [-w] [-i] [-l] [-h]
-n: Name of the application to be used in column headers.
-c: Column header name for each parameter.
You should give multiple header names in order for each directory in the results directory structure.
Default: Heap Size,Concurrent Users,Message Size (Bytes),Back-end Service Delay (ms)
-r: Regular expression with a single group to extract parameter value from directory name.
You should give multiple regular expressions in order for each directory in the results directory structure.
Default: ([0-9]+[a-zA-Z])_heap,([0-9]+)_users,([0-9]+)B,([0-9]+)ms_sleep
-x: Print column names and exit.
-p: Prefix of the files to get metrics (Load Average, GC, etc).
-g: Path of GCViewer Jar file, which will be used to analyze GC logs.
-d: Results directory. Default ./jmeter/results.
-j: Number of JMeter servers. If n=1, only client was used. If n > 1, remote JMeter servers were used. Default 1.
-k: Number of Application instances. Default 1.
-w: Use warmup results instead of measurement results.
-i: Include GC statistics and load averages for other servers.
-l: Exclude Netty Back-end Service statistics. Works with -i.
-h: Display this help and exit.
Use csv-to-markdown-converter.py
to convert CSV results into Markdown format.
ubuntu@server:~$ ./jmeter/csv-to-markdown-converter.py
Usage: {Input File(.csv)} {Output File (.md)}
Use create-summary-markdown.py
to create a markdown file, which will be a summary of performance test results. This scripts
looks for a Jinja2
template with the name summary.md
in templates
directory.
ubuntu@server:~$ ./jmeter/create-summary-markdown.py -h
usage: create-summary-markdown.py [-h] --json-files JSON_FILES
[JSON_FILES ...] --column-names COLUMN_NAMES
[COLUMN_NAMES ...]
Create summary report
optional arguments:
-h, --help show this help message and exit
--json-files JSON_FILES [JSON_FILES ...]
JSON files with parameters.
--column-names COLUMN_NAMES [COLUMN_NAMES ...]
Columns to include in the report.
The "jtl-splitter" directory has a Java program to split a single JTL file into warmup and measurement based on the number of minutes given as the warmup time.
When reporting the results for the performance tests, some specified number of minutes from the beginning of the test are considered as the "Java Warm-up Time" and the from the final results, the warm-up duration is excluded. By doing this, the results reported from the test will only consider the steady-state of the server.
This program should be invoked by the performance testing script after completing the JMeter performance test.
For example if you specify 5 minutes warmup-time, the JTL splitter splits the results.jtl
file and the results-warmup.jtl
file will have the test results for first 5 minutes. The results after 5 minutes will be in results-measurement.jtl
.
ubuntu@server:~$ ./jtl-splitter/jtl-splitter.sh -h
Usage:
./jtl-splitter/jtl-splitter.sh [-m <heap_size>] [-h] -- [jtl_splitter_flags]
-m: The heap memory size. Default: 1g
-h: Display this help and exit.
JTL Splitter usage:
ubuntu@server:~$ ./jtl-splitter/jtl-splitter.sh -- -h
Usage: JTLSplitter [options]
Options:
-d, --delete-jtl-file-on-exit
Delete JTL File on exit
Default: false
-h, --help
Display Help
* -f, --jtlfile
JTL File
-n, --precision
Precision to use in statistics
Default: 2
-p, --progress
Show progress
Default: false
-s, --summarize
Summarize results
Default: false
-u, --time-unit
Time Unit
Default: MINUTES
Possible Values: [NANOSECONDS, MICROSECONDS, MILLISECONDS, SECONDS, MINUTES, HOURS, DAYS]
* -t, --warmup-time
Warmup Time
Default: 0
The "netty-service" directory has a simple Netty HTTP Echo Service, which will echo back the body data in the HTTP request.
The Netty HTTP Echo Service should be started by the performance testing script.
ubuntu@server:~$ ./netty-service/netty-start.sh -h
Usage:
./netty-service/netty-start.sh [-m <heap_size>] [-h] -- [netty_service_flags]
-m: The heap memory size of Netty Service. Default: 4g
-h: Display this help and exit.
The script also accepts an argument to specify the number of milliseconds to sleep before sending response. This is useful to test the performance with delays.
ubuntu@server:~$ ./netty-service/netty-start.sh -- -h
Starting Netty
Usage: EchoHttpServer [options]
Options:
--boss-threads
Boss Threads
Default: 4
--delay
Response delay in milliseconds
Default: 0
-h, --help
Display Help
--http2
Use HTTP/2 protocol instead of HTTP/1.1
Default: false
--key-store-file
Keystore file
--key-store-password
Keystore password
Default: <empty string>
--port
Server Port
Default: 8688
--ssl
Enable SSL
Default: false
--worker-threads
Worker Threads
Default: 8
The "payloads" directory has a Java program to generate JSON payloads with different sizes.
By default, the script generates 50B, 1KiB, 10KiB, and 100KiB JSON files.
If you want to generate different payload sizes, pass the payload sizes as parameters.
The performance testing script can call this script to generate payloads required for the performance test.
ubuntu@server:~$ ./payloads/generate-payloads.sh -h
Usage:
./payloads/generate-payloads.sh [-p <payload_type>] [-s <payload_size>]
-p: The Payload Type.
-s: The Payload Size. You can give multiple payload sizes.
-h: Display this help and exit.
The "sar" directory has a simple script to install System Activity Report (SAR) in Linux and configure it to run every one minute.
The script needs to be run as root.
ubuntu@server:~$ sudo ./sar/install-sar.sh -h
Usage:
./sar/install-sar.sh [-h]
-h: Display this help and exit.
The "setup" directory has the scripts to setup instances (for example, JMeter Client, JMeter Server, Netty Server, etc.)
The setup-common.sh
script is used by all other setup scripts to do some common operations.
ubuntu@server:~$ sudo ./setup/setup-common.sh -h
Usage:
./setup/setup-common.sh [-g] [-p <package>] [-w <url_to_download>] [-o <output_name>]
-g: Upgrade distribution
-p: Package to install. You can give multiple -p options.
-w: Download URLs. You can give multiple URLs to download.
-o: Output name of the downloaded file. You can give multiple names for a given set of URLs.
-h: Display this help and exit.
The setup-jmeter.sh
installs JMeter and JMeter plugins. This script uses the install-jmeter.sh
script in "jmeter"
directory.
ubuntu@server:~$ sudo ./setup/setup-jmeter.sh -h
Usage:
./setup/setup-jmeter.sh -i <installation_dir> [-j <jmeter_plugin>] [-g] [-p <package>] [-w <url_to_download>] [-o <output_name>]
-g: Upgrade distribution
-p: Package to install. You can give multiple -p options.
-w: Download URLs. You can give multiple URLs to download.
-o: Output name of the downloaded file. You can give multiple names for a given set of URLs.
-i: The JMeter installation directory.
-j: The JMeter plugin name. You can give multiple JMeter plugins to install.
-h: Display this help and exit.
The setup-jmeter-client.sh
uses setup-jmeter.sh
internally. It also creates the SSH configurations to execute commands
in other instances.
ubuntu@server:~$ sudo ./setup/setup-jmeter-client.sh -h
Usage:
./setup/setup-jmeter-client.sh -k <key_file> -i <installation_dir> -c <ssh_config_location> -a <ssh_alias> -n <ssh_hostname> [-j <jmeter_plugin>] [-g] [-p <package>] [-w <url_to_download>] [-o <output_name>]
-g: Upgrade distribution
-p: Package to install. You can give multiple -p options.
-w: Download URLs. You can give multiple URLs to download.
-o: Output name of the downloaded file. You can give multiple names for a given set of URLs.
-k: The key file location.
-i: The JMeter installation directory.
-c: The SSH config location.
-a: SSH Alias. You can give multiple ssh aliases.
-n: SSH Hostname. You can give multiple ssh hostnames for a given set of ssh aliases.
-j: The JMeter plugin name. You can give multiple JMeter plugins to install.
-h: Display this help and exit.
The setup-netty.sh
installs Netty Server and OpenJDK
ubuntu@server:~$ sudo ./setup/setup-netty.sh -h
Usage:
./setup/setup-netty.sh [-g] [-p <package>] [-w <url_to_download>] [-o <output_name>]
-g: Upgrade distribution
-p: Package to install. You can give multiple -p options.
-w: Download URLs. You can give multiple URLs to download.
-o: Output name of the downloaded file. You can give multiple names for a given set of URLs.
-h: Display this help and exit.
Copyright 2017 WSO2 Inc. (http://wso2.com)
Licensed under the Apache License, Version 2.0