Taurus > User Manual > JMeter
fork me on GitHub
Automation-friendly framework for Continuous Testing



JMeter Executor

This executor type is used by default, it uses Apache JMeter as underlying tool.

JMeter Location & Auto-Installation

If there is no JMeter installed at the configured path, Taurus will attempt to install latest JMeter and Plugins into this location, by default ~/.bzt/jmeter-taurus/{version}/bin/jmeter. You can change this setting to your preferred JMeter location (consider putting it into ~/.bzt-rc file). All module settings that relates to JMeter path and auto-installing are listed below:

modules:
  jmeter:
    path: ~/.bzt/jmeter-taurus/bin/jmeter
    download-link: https://archive.apache.org/dist/jmeter/binaries/apache-jmeter-{version}.zip
    version: 3.0  # minimal supported version of JMeter is 2.9
    force-ctg: true   # true by default
    detect-plugins: true
    plugins:
    - jpgc-json=2.2
    - jmeter-ftp
    - jpgc-casutg

force-ctg allows you to switch off the usage of ConcurrentThreadGroup for jmx script modifications purpose. This group provide steps execution parameter but requires Custom Thread Groups plugin (installed by default)

With version parameter you can ask for specific tool version or use autodetect with auto value. In that case taurus will analyze content of jmx file and try to guess appropriate the JMeter version.

plugins option lets you describe list of JMeter plugins you want to use. If plugins option isn't found only following plugins will be installed: jpgc-casutg, jpgc-dummy, jpgc-ffw, jpgc-fifo, jpgc-functions, jpgc-json, jpgc-perfmon, jpgc-prmctl, jpgc-tst. Keep in mind: you can change plugins list only for clean installation. If you already have JMeter placed at path you need to remove it for plugins installation purpose.

JMeter Plugins Manager allows you to install necessary plugins for your jmx file automatically and this feature doesn't require clean installation. You can turn it off with detect-plugins option. If you use your own installation of JMeter (with path option) make sure it includes jmeter-plugins-manager 0.16 or newer.

Run Existing JMX File

execution:
- scenario: simple

scenarios:
  simple:
    script: tests/jmx/dummy.jmx

or simply bzt tests/jmx/dummy.jmx

TODO: explain how multi-thread group will accept concurrency with maintained proportion

JMeter Properties and Variables

There are two places to specify JMeter properties: global at module-level and local at scenario-level. Scenario properties are merged into global properties and resulting set comes as input for JMeter, see corresponding .properties file in artifacts. You may also specify system properties for JMeter in system-properties section. They comes as system.properties file in artifacts.

Global properties are set like this:

modules:
  jmeter:
    properties:
      my-hostname: www.pre-test.com
      log_level.jmeter: WARN
      log_level.jmeter.threads: DEBUG
    system-properties:
      sun.net.http.allowRestrictedHeaders: "true"

Scenario-level properties are set like this:

scenarios:
  prop_example: 
    properties:
        my-hostname: www.prod.com
        log_level.jmeter: DEBUG

You can use properties in different parts of JMeter parameters to tune your script behaviour:

execution:
- concurrency: ${__P(my_conc,3)}    # use `my_conc` prop or default=3 if property isn't found
  ramp-up: 30
  hold-for: ${__P(my_hold,10)}
  scenario: with_prop

modules:
  jmeter:
    properties:
      my_conc: 10
      my_hold: 20

scenarios:
  with_prop:
    requests:
    - http://blazedemo.com/${__P(sub_dir)}
    - https://blazemeter.com/${__P(sub_dir)}
    properties:
      my_hold: 15   # scenario-level property has priority
      sub_dir: contacts

Usage of variables are similar but they can be used on scenario level only:

scenarios:
  sc_with_vars:
    variables:
      subdir: contacts
      ref: http://gettaurus.org
    requests:
    - url: http://blazedemo.com/${subdir}
      headers:
        Referer: ${ref}
    - url: https://blazemeter.com/${subdir}
      headers:
        Referer: ${ref}

Open JMeter GUI

When you want to verify or debug the JMX file that were generated from your requests scenario, you don't need to search for the file on disk, just enable GUI mode for JMeter module:

modules:
  jmeter:
    gui: false  # set it to true to open JMeter GUI instead of running non-GUI test

For the command-line, use alias -gui or option -o modules.jmeter.gui=true, without the need to edit configuration file.

Run JMeter in Distributed Mode

Distributed mode for JMeter is enabled with simple option distributed under execution settings, listing JMeter servers under it:

execution:
- distributed: 
  - host1.mynet.com
  - host2.mynet.com
  - host3.mynet.com
  scenario: some_scenario
  
scenarios:
  some_scenario:
    script: my-test.jmx

For accurate load calculation don't forget to choose different hostname values for slave hosts. If you have any properties specified in settings, they will be sent to remote nodes.

Shutdown Delay

By default, Taurus tries to call graceful JMeter shutdown by using its UDP shutdown port (this works only for non-GUI). There is option to wait for JMeter to exit before killing it forcefully, called shutdown-wait. By default, its value is 5 seconds.

Modifications for Existing Scripts

JMeter executor allows you to apply some modifications to the JMX file before running JMeter (this affects both existing JMXes and generated from requests):

scenarios:
  modification_example:
    script: tests/jmx/dummy.jmx
    modifications:
      disable:  # Names of the tree elements to disable
      - Thread Group 1
      enable:  # Names of the tree elements to enable
      - Thread Group 2
      set-prop:  # Set element properties, selected as [Element Name]>[property name]
        "HTTP Sampler>HTTPSampler.connect_timeout": "0"
        "HTTP Sampler>HTTPSampler.protocol": "https"

If selector for set-prop isn't found, taurus tries to create stringProp jmx element with last element of selector as name and sets given value for it. So you can create simple properties in jmx if it's necessary.

Building Test Plan from Config

Scenario that has requests element makes Taurus to generate the script for underlying tools automatically. For now, this is available for JMeter and partially available for some other tools.

The requests element must contain a list of requests, each with its settings and child elements (assertions, extractors). Also there are additional configuration elements for requests-based scenario, described below.

Scenario is the sequence of steps and some settings that will be used by underlying tools (JMeter, Grinder, Gatling) on execution stage.

Scenarios are listed in top-level scenarios element and referred from executions by their alias:

scenarios:
  get-requests:  # the alias for scenario
    requests:
    - http://localhost/1
    - http://localhost/2

execution:
- scenario: get-requests  # alias from above is used

Global Settings

Scenario has some global settings:

scenarios:
  get-requests:  
    store-cache: true  # browser cache simulation, enabled by default
    store-cookie: true  # browser cookies simulation, enabled by default
    headers: # global headers
      header-name: header-value
    think-time: 1s500ms  # global delay between each request
    timeout: 500ms  #  timeout for connecting, receiving results
    default-address: "https://www.blazedemo.com:8080"  # http request defaults scheme, domain, port
    keepalive: true  # true by default, applied on all requests in scenario
    retrieve-resources: true  # true by default, retrieves all embedded resources from HTML pages
    retrieve-resources-regex: ^((?!google|facebook).)*$  # regular expression used to match any resource
                                                         # URLs found in HTML document against. Unset by default
    concurrent-pool-size: 4  # concurrent pool size for resources download, 4 by default
    use-dns-cache-mgr: true  # use DNS Cache Manager to test resources
                             # behind dns load balancers. True by default.
    force-parent-sample: true  # generate only parent sample for transaction controllers.
                               # True by default
    content-encoding: utf-8  # global content encoding, applied to all requests.
                             # Unset by default
    follow-redirects: true  # follow redirects for all HTTP requests
    random-source-ip: false  # use one of host IPs to send requests, chosen randomly.
                             # False by default
    data-sources: # list of external data sources
    - path/to/my.csv  # this is a shorthand form
    - path: path/to/another.csv  # this is full form, path option is required
      delimiter: ';'  # CSV delimiter, auto-detected by default
      quoted: false  # allow quoted data
      loop: true  # loop over in case of end-of-file reached if true, stop thread if false
      variable-names: id,name  # delimiter-separated list of variable names, empty by default

Note that timeout also sets duration assertion that will mark response failed if response time was more than timeout.

If you want to use JMeter properties in default-address, you'll have to specify mandatory scheme and separate address/port. Like this: default-address: https://${__P(hostname)}:${__P(port)}.

Requests

Request objects can be of two kinds:

  1. Plain HTTP requests
  2. Logic blocks that allow user to control execution flow of test session.

HTTP Requests

The base element for requests scenario is HTTP Request. In its simplest form it contains just the URL as string:

scenarios:
  get-requests:  
    requests:
    - http://localhost/1
    - http://localhost/2

The full form for request is dictionary, all fields except url are optional:

scenarios:
  my-req: 
    requests:
    - url: http://blazedemo.com/  # url to hit
      method: GET  # request method (GET, POST, PUT, DELETE)
      label: homepage  # sampler label

      body: 'request-body-string'  # if present, will be used as body 
      body:  # generate query string based on parameters and request type
        param1: value1
        param2: value2
      body-file: path/to/file.txt  # this file contents will be used as post body

      upload-files:  # attach files to form (and enable multipart/form-data)
      - param: summaryReport  # form parameter name
        path: report.pdf  # path to file
        mime-type: application/pdf  # optional, Taurus will attempt to guess it automatically

      headers:  # local headers that override global
        Authentication: Token 1234567890
        Referer: http://taurus.blazemeter/docs
      think-time: 1s  # local think-time, overrides global
      timeout: 1s  # local timeout, overrides global
      content-encoding: utf-8  # content encoding (at JMeter's level), unset by default
      follow-redirects: true  # follow HTTP redirects
      random-source-ip: false  # use one of host IPs to send the request (chosen randomly).
                               # False by default

      extract-regexp: {}  # explained below
      extract-jsonpath: {}  # explained below
      assert: []  # explained below
      jsr223: []  # explained below

Notes for upload-files:

  • POST request method requires non-empty param values
  • PUT method allows only one file in upload-files block
Extractors

Extractors are the objects that attached to request to take a piece of the response and use it in following requests. The concept is based on JMeter's extractors. The following types of extractors are supported:

  • by regular expression
  • by JSONPath expression
  • by CSS/JQuery selectors
  • by XPath query

To specify extractors in shorthand form, use following configuration:

scenarios:
  my-req: 
    requests:
    - url: http://blazedemo.com/  
      extract-regexp: # dictionary under it has form <var name>: <regular expression>
        page_title: <title>(\w+)</title>  #  must have at least one capture group
      extract-jsonpath: # dictionary under it has form <var name>: <JSONPath expression>
        varname: $.jsonpath[0].expression
    - url: http://blazedemo.com/${varname_1}/${page_title}  # that's how we use those variables
      extract-css-jquery: # dictionary under it has form <var name>: <CSS/JQuery selector>
        extractor1: input[name~=my_input]
    - url: http://blazedemo.com/${varname}/${extractor1}.xml
      extract-xpath:
        title: /html/head/title

The full form for extractors is:

scenarios:
  my-req: 
    requests:
    - url: http://blazedemo.com/  
      extract-regexp:
        page_title:
          regexp: <title>(\w+)</title>  # regular expression
          default: NOT_FOUND  # default value to use when regexp not found
          match-no: 1  # if multiple values has matched, which match use (0=random)
          template: 1  # which capture group to take, integer or template string
          subject: body  #  subject for search
      extract-jsonpath:   
        varname:
          jsonpath: $.jsonpath[0]  # jsonpath expression
          default: NOT_FOUND  # default value to use when jsonpath not found
          from-variable: JM_VAR # JMeter variable for search
          concat: false # \
          scope: all    # - see below
          match-no: 4   # /
    - url: http://blazedemo.com/${varname}/${page_title}
      extract-css-jquery:
        extractor2:
          expression: input[name=JMeter]
          attribute: value
          match-no: 1
          default: NOT_FOUND
    - url: http://blazedemo.com/${varname}/${extractor2}.xml
      extract-xpath:
        destination:
          xpath: /order/client/address
          default: NOT_FOUND
          validate-xml: false
          ignore-whitespace: true
          match-no: -1
          use-namespaces: false
          use-tolerant-parser: false

Parameters of jsonpath exractor concat, scope and match-num work only on JMeter >= 3.0 If several results are found they will be concatenated with ',' if concat. You can choose scope for applying expressions. Possible targets are:

  • all - main sample and sub-samples
  • children - sub-samples
  • variable for search in JMeter variables

Default value of scope is empty, it means search in main sample only.

match-no allows to choose the specific result from several ones. Default value is -1 - generation of variables varname_1, varname_2, etc. It means if you ask for some_var_name JMeter won't generate variable with exactly that name by default. Set match-no to another value if you want to change this behaviour.

Possible subjects for regexp are:

  • body
  • headers
  • http-code
  • url
Assertions

Assertions are attached to request elements and used to set fail status on the response. Fail status for the response is not the same as response code for JMeter. Currently three types of response assertions are available.

First one checks http response fields, its short form looks like this:

scenarios:
  my-req: 
    requests:
    - url: http://blazedemo.com/  
      assert:  # contains list of regular expressions to check
      - .+App.+

The full form has following format:

scenarios:
  my-req: 
    requests:
    - url: http://blazedemo.com/  
      assert:
       - contains:  # list of strings to check
         - .+App.+ 
         subject: body  # subject for search
         regexp: true  # treat string as regular expression
         not: false  # invert condition - fail if found
         assume-success: false  # mark sample successful before asserting it

Possible subjects are:

  • body
  • headers
  • http-code

The second assertion type is used to perform validation of JSON response against JSONPath expression.

scenarios:
  my-req:
    requests:
    - url: http://blazedemo.com/
      assert-jsonpath:  # contains list of options
        - "$."  # if this JSONPath not found, assert will fail
        - "$.result[0]" # there can be multiple JSONPaths provided

Full form:

scenarios:
  my-req:
    requests:
    - url: http://blazedemo.com/
      assert-jsonpath:
      - jsonpath: "$." # path to value, validation fails if path not exists
        validate: true # validate against expected value
        expected-value: "value" # the value we are expecting to validate, default: false
        regexp: true  # if the value is regular expression, default: true
        expect-null: false  # expected value is null
        invert: false # invert condition

And the third assertion type uses XPath query to validate XML response.

scenarios:
  assertion-demo:
    requests:
    - url: http://blazedemo.com/
      assert-xpath:  # contains list of xpath queries
        - "/bookstore/book"  # if this XPath won't be matched, assert will fail
        - "/html/head/title" # you can provide multiple XPath queries

Full form:

scenarios:
  my-req:
    requests:
    - url: http://blazedemo.com/
      assert-xpath:
      - xpath: "/html/head/title/text()='My title'" # query that compares XPath query result with some value
        use-tolerant-parser: false  # use error-tolerant XML parser
        ignore-whitespace: true # ignore whitespaces in XML (has no effect when `use-tolerant-parser` is true)
        validate: false # validate XML against its schema (has no effect when `use-tolerant-parser` is true)
        invert: false # invert condition
JSR223 Blocks

Sometimes you may want to use a JSR223 Pre/Post Processor to execute a code block before or after each request. Taurus allows that with jsr223 block.

Minimal example that will generate one JSR223 Post Processor.

scenarios:
  jsr-example:
    requests:
    - url: http://blazedemo.com/
      jsr223: 'vars.put("varname", "somevalue")'  # inline script to execute, unless script-file is specified

The example above uses defaults and inline script. If you want to use language different from groovy or use separate script file, please use extended form of jsr223 with key-value options.

Each jsr223 element can define the following fields:

  • language - script language ('beanshell', 'bsh', 'ecmascript', 'groovy', 'java', 'javascript', 'jexl', 'jexl2')
  • script-file - path to script file
  • script-text - inline code, specified directly in config file
  • parameters - string of parameters to pass to script, empty by default
  • execute - whether to execute script before or after the request

If execute field is set to after - Taurus will generate a JSR223 PostProcessor, if set to before - a PreProcessor. By default it's set to after.

Full form:

scenarios:
  jsr-example:
    requests:
    - url: http://blazedemo.com/
      jsr223:
      - language: javascript
        script-file: preproc.js
        parameters: foo bar
        execute: before
      - language: beanshell
        script-file: postproc.bsh
        execute: after

Logic Blocks

Taurus allows to control execution flow with the following constructs:

  • if blocks
  • loop blocks
  • while blocks
  • foreach blocks
  • transaction blocks
  • include-scenario blocks
  • action blocks
If Blocks

if blocks allow conditional execution of HTTP requests.

Each if block should contain a mandatory then field, and an optional else field. Both then and else fields should contain lists of requests.

Here's a simple example:

scenarios:
  if_example:
    variables:
      searchEngine: google
    requests:
    - if: '"${searchEngine}" == "google"'
      then:
        - https://google.com/
      else:
        - https://bing.com/

Note that Taurus compiles if blocks to JMeter's If Controllers, so <condition> must be in JMeter's format.

Logic blocks can also be nested:

scenarios:
  nested_example:
    requests:
    - if: <condition1>
      then:
      - if: <condition2>
        then:
        - https://google.com/
        else:
        - https://yahoo.com/
      else:
      - https://bing.com/

And here's the real-world example of using if blocks:

scenarios:
  complex:
    requests:
    # first request is a plain HTTP request that sets `status_code`
    # and `username` variables
    - url: https://api.example.com/v1/media/search
      extract-jsonpath:
        status_code: $.meta.code
        username: $.data.[0].user.username
 
    # branch on `status_code` value
    - if: '"${status_code}" == "200"'
      then:
        - https://example.com/${username}
Loop Blocks

loop blocks allow repeated execution of requests. Nested requests are to be specified with do field.

scenarios:
  loop_example:
    requests:
    - loop: 10
      do:
      - http://blazedemo.com/

If you want to loop requests forever, you can specify string forever as loop value.

scenarios:
  forever_example:
    requests:
    - loop: forever
      do:
      - http://blazedemo.com/

Note that loop blocks correspond to JMeter's Loop Controllers.

While Blocks

while block is similar to while loops in many programming languages. It allows conditional repeated execution of requests. while blocks are compiled to JMeter's While Controllers.

scenarios:
  while_example:
    requests:
    - while: ${JMeterThread.last_sample_ok}
      do:
      - http://blazedemo.com/
Foreach Blocks

foreach blocks allow you to iterate over a collection of values. They are compiled to JMeter ForEach Controllers.

Syntax:

requests:
- foreach: <elementName> in <collection>
  do:
  - http://${elementName}/

Concrete example:

scenarios:
  complex_foreach:
    requests:
    - url: https://api.example.com/v1/media/search
      extract-jsonpath:
        usernames: $.data.[:100].user.username  # grab first 100 usernames
    - foreach: name in usernames
      do:
      - https://example.com/user/${name}
Transaction Blocks

transaction blocks allow wrapping http requests in a transaction. transaction blocks correspond to JMeter's Transaction Controllers.

Example:

scenarios:
  transaction_example:
    requests:
    - transaction: Customer Session
      force-parent-sample: false  # True by default
      include-timers: true  # add timers and pre-/post-processors execution time to samples
      do:
      - http://example.com/shop
      - http://example.com/shop/items/1
      - http://example.com/shop/items/2
      - http://example.com/card
      - http://example.com/checkout

Take note: you can specify force-parent-sample on both levels - scenario and transaction. If both are found local (transaction) value has priority.

Include Scenario Blocks

include-scenario block allows you to include scenario in another one. You can use it to split your test plan into a few of independent scenarios that can be reused.

Example:

scenarios:
  login:
    data-sources:
    - logins.csv
    requests:
    - url: http://example.com/login
      method: POST
      body:
        user: ${username}
        password: ${password}
  logout:
    requests:
    - url: http://example.com/logout
      method: POST
  shop-session:
    requests:
    - include-scenario: login
    - http://example.com/shop/items/1
    - http://example.com/checkout
    - include-scenario: logout

Taurus translates each include-scenario block to a JMeter's Simple Controller and puts all scenario-level settings and requests there. Keep in mind: the following scenario-level parameters of including scenario have no effect for included ones:

  • keepalive
  • timeout
  • think-time
  • follow-redirects
  • properties
Action Blocks

action block allows you to specify a thread-specific action that will be performed. You can use it to pause or stop the current thread, or force it to go to the next loop iteration.

The following actions are available:

  • pause - pause the target thread (pause duration is controlled with pause-duration field)
  • stop - stop the target thread gracefully
  • stop-now - stop the test without waiting for samples to complete
  • continue - send target thread to the next iteration of the loop

Actions can be applied to the following targets:

  • current-thread - set by default
  • all-threads - action will be applied to all threads (unavailable for continue action)

Examples:

scenarios:
  action_example:
    requests:
    - action: pause
      target: current-thread
      pause-duration: 1s500ms
    - action: stop-now
      target: all-threads
Set Variables Blocks

set-variables block allows you to set JMeter variables from other variables.

Example:

scenarios:
  set_vars_example:
    variables:
      foo: BAR
    requests:
    - http://blazedemo.com/?foo=${foo}
    - set-variables:
        foo: BAZ

This example will set initial value of ${foo} to be "BAR", but after first iteration it will be changed to "BAZ".

You can consider this block to be a syntactic sugar over JSR223 blocks, because that's exactly how it works.

User cookies

Taurus allows you to set up some user cookies with follow syntax:

scenarios:
  little_cookies:
    requests:
    - http://blazedemo.com/login
    - https://blazemeter.com/pricing
    cookies:
    - name: n0  # name of cookie, required
      value: v0 # value of cookie, required
      domain: blazedemo.com # hosts to which the cookie will be sent, required
    - name: n1
      value: v1
      domain: blazemeter.com
      path: /pricing  # must exist in the request for sending cookie
      secure: true    # send only through https, optional, defalut: false

JMeter Test Log

You can tune JTL file content with option write-xml-jtl. Possible values are 'error' (default), 'full', or any other value for 'none'. Keep in mind: max full logging can seriously load your system.

execution:
- write-xml-jtl: full
  scenario: simple_script
  
scenarios:
  simple_script:
    script: my.jmx
 

Another way to adjust verbosity is to change flags in xml-jtl-flags dictionary. Next example shows all flags with default values (you don't have to use full dictionary if you want to change some from them):

modules:
  jmeter:
    xml-jtl-flags:
      xml: true
      fieldNames: true
      time: true
      timestamp: true
      latency: true
      connectTime: true
      success: true
      label: true
      code: true
      message: true
      threadName: true
      dataType: true
      encoding: true
      assertions: true
      subresults: true
      responseData: false
      samplerData: false
      responseHeaders: true
      requestHeaders: true
      responseDataOnError: true
      saveAssertionResultsFailureMessage: true
      bytes: true
      threadCounts: true
      url: true

Remember: some logging information might be used by assertions so change log verbosity can affect them.

JMeter JVM Memory Limit

You can tweak JMeter's memory limit (aka, -Xmx JVM option) with memory-xmx setting. Use K, M or G suffixes to specify memory limit in kilobytes, megabytes or gigabytes.

Example:

modules:
  jmeter:
    memory-xmx: 4G  # allow JMeter to use up to 4G of memory

SoapUI Integration

You can specify SoapUI projects in place of JMX scripts for JMeter executor. The executor will convert them into Taurus scenarios and execute them with JMeter.

Example:

execution:
- concurrency: 10
  hold-for: 5m
  scenario: soapui-project

scenarios:
  soapui-project:
    script: project.xml
    test-case: TestIndex

You can read more on that here.