Skip to content

Latest commit

 

History

History
110 lines (85 loc) · 8.23 KB

README.md

File metadata and controls

110 lines (85 loc) · 8.23 KB
Raw

Elasticsearch For Beginners: Generate and Upload Randomized Test Data

Because everybody loves test data.

Ok, so what is this thing doing?

es_test_data.py lets you generate and upload randomized test data to your ES cluster so you can start running queries, see what performance is like, and verify your cluster is able to handle the load.
It allows for easy configuring of what the test documents look like, what kind of data types they include and what the field names are called.

Cool, how do I use this?

Prerequisites

Let's assume you have an Elasticsearch cluster running. If not, set it up locally and point your browser to http://localhost:9200 to see if it's up.

Python and Tornado and NumPy are used, run pip install tornado numpy to install Tornado and NumPy if you don't have it already.

Lets get started

It's as simple as this:

$ python es_test_data.py --es_url=http://localhost:9200
[I 150604 15:43:19 es_test_data:42] Trying to create index http://localhost:9200/test_data
[I 150604 15:43:19 es_test_data:47] Guess the index exists already
[I 150604 15:43:19 es_test_data:184] Generating 10000 docs, upload batch size is 1000
[I 150604 15:43:19 es_test_data:62] Upload: OK - upload took:    25ms, total docs uploaded:    1000
[I 150604 15:43:20 es_test_data:62] Upload: OK - upload took:    25ms, total docs uploaded:    2000
[I 150604 15:43:20 es_test_data:62] Upload: OK - upload took:    19ms, total docs uploaded:    3000
[I 150604 15:43:20 es_test_data:62] Upload: OK - upload took:    18ms, total docs uploaded:    4000
[I 150604 15:43:20 es_test_data:62] Upload: OK - upload took:    27ms, total docs uploaded:    5000
[I 150604 15:43:20 es_test_data:62] Upload: OK - upload took:    19ms, total docs uploaded:    6000
[I 150604 15:43:20 es_test_data:62] Upload: OK - upload took:    15ms, total docs uploaded:    7000
[I 150604 15:43:20 es_test_data:62] Upload: OK - upload took:    24ms, total docs uploaded:    8000
[I 150604 15:43:20 es_test_data:62] Upload: OK - upload took:    32ms, total docs uploaded:    9000
[I 150604 15:43:20 es_test_data:62] Upload: OK - upload took:    31ms, total docs uploaded:   10000
[I 150604 15:43:20 es_test_data:216] Done - total docs uploaded: 10000, took 1 seconds
[I 150604 15:43:20 es_test_data:217] Bulk upload average:           23 ms
[I 150604 15:43:20 es_test_data:218] Bulk upload median:            24 ms
[I 150604 15:43:20 es_test_data:219] Bulk upload 95th percentile:   31 ms
$

Without any command line options, it will generate and upload 1000 documents of the format

{
    "name":<<str>>,
    "age":<<int>>,
    "last_updated":<<ts>>
}

to an Elasticsearch cluster at http://localhost:9200 to an index called test_data.

Not bad but what can I configure?

python es_test_data.py --help gives you the full set of command line options, here are the most important ones:

  • --es_url=http://localhost:9200 the base URL of your ES node, don't include the index name
  • --count=### number of documents to generate and upload
  • --index_name=test_data the name of the index to upload the data to. If it doesn't exist it'll be created with these options
    • --num_of_shards=2 the number of shards for the index
    • num_of_replicas=0 the number of replicas for the index
  • --batch_size=### we use bulk upload to send the docs to ES, this option controls how many we send at a time
  • --force_init_index=False if True it will delete and re-create the index
  • --dict_file=filename.dic if provided the dict data type will use words from the dictionary file, format is one word per line. The entire file is loaded at start-up so be careful with (very) large files. You can download wordlists e.g.. from here.
  • --cities_file=filename.cvs if provided the cities will be loaded from the CSV file. Default is worldcities.csv which can be downloaded from here.
  • --num_of_cities if provided, sets the number of cities to use when generating city points. Default is to use all cities loaded via --cities_file.

What about the document format?

Glad you're asking, let's get to the doc format.
The doc format is configured via --format=<<FORMAT>> with the default being name:str,age:int,last_updated:ts.

The general syntax looks like this:

<<field_name>>:<<field_type>>,<<field_name>>::<<field_type>>, ...

For every document, es_test_data.py will generate random values for each of the fields configured.

Currently supported field types are:

  • bool returns a random true or false
  • ts:min_days:max_days a timestamp (in milliseconds), randomly picked between now - min_days and now+max_days. Defaults to +/- 30. If you wanted just days in the past you would put zero for max_days (e.g. ts:30:0)
  • ts_series:min_days:max_days:delta:interval a timestamp (in milliseconds) starting between now-min_days and now+max_days incremented delta ms each data point. Starts over after interval number of points
  • ipv4 returns a random ipv4
  • tstxt a timestamp in the "%Y-%m-%dT%H:%M:%S.000-0000" format, randomly picked between now +/- 30 days
  • int:min:max a random integer between min and max. If min and max` are not provided they default to 0 and 100000
  • str:min:max a word ( as in, a string), made up of min to max random upper/lowercase and digit characters. If min and max are optional, defaulting to 3 and 10
  • str_series:min:max:interval a word ( as in, a string), made up of min to max random upper/lowercase and digit characters. If min and max are optional, defaulting to 3 and 10. Stay the same for inerval number of documents
  • keyword:min:max a word ( as in, a string), made up of min to max random upper/lowercase and digit characters. If min and max are optional, defaulting to 3 and 10
  • words:min:max a random number of strs, separated by space, min and max are optional, defaulting to '2' and 10
  • dict:min:max a random number of entries from the dictionary file, separated by space, min and max are optional, defaulting to '2' and 10
  • text:words:min:max a random number of words seperated by space from a given list of - seperated words, the words are optional defaulting to text1 text2 and text3, min and max are optional, defaulting to 1 and 1
  • geo_point:min_lat:max_lat:min_lon:max_lon a random geopoint between min_lat, max_lat, min_lon, and max_lon. If min and max values are not provided the default to the entire world
  • ellipse:major_mean:minor_mean:major_std:minor:std:num_points a random ellipse of random size and tilt based at a random location based on mean and standard deviation provided. Ellipse is drawn as a polygon with num_points verticies
  • cities:min_rad:max_rad return a random geopoint within min_rad and max_rad meters from a chosen random city loaded via --cities_file.
  • cities_path_series:length:min_rad:max_rad:heading_std:speed_start:speed_std:interval:interval_std Creates a series of geo_points of length starting at a random geopoint within min_rad and max_rad meters from a chosen random city loaded via --cities_file. Path starts at a random heading and varies with heading_std and has a starting speed_start (m/s) varying with speed_std. A new point is created every interval seconds but varies with interval_std
  • ellipse_cities:major_mean:minor_mean:major_std:minor:std:num_points:sigma_degrees a random ellipse of random size and tilt based near a random city based on mean and standard deviation provided. Ellipse is drawn as a polygon with num_points verticies. Centers are a normial distribution away from city center with sigma_degree std dev.
  • path:num_points:heading_std:speed_start:speed:std creates a path of num_points long that starts at a random points on a random heading. It changes heading based on a normal distribution with heading_std as the standard deviation. It starts at speed_start (m/s) and changes based on a normal distribution with speed_std as the standard deviation.

todo

  • document the remaining cmd line options
  • more different format types
  • ...

All suggestions, comments, ideas, pull requests are welcome!