influxdb/cmd/benchmark/benchmark_config.sample.toml

# This is a configuration file for the InfluxDB benchmarking tool.

# High level stats for the benchmark run will be reported to this log file.
log_file = "benchmark.log"

# Output summary stats after each block of this number of points has been posted.
output_after_count = 10000

# Stats for the benchmark run will be reported to this InfluxDB server. Track it over time, friend.
[stats_server]
connection_string = "localhost:8086"
database = "reports"
user = "user"
password = "pass"
is_secure = false
skip_verify = false
timeout = "10s"

# A regular database, user, and password to read and write data on the cluster being benchmarked.
[cluster_credentials]
database = "benchmark"
user = "paul"
password = "pass"

[load_settings]
# The benchmarker will make this many connections per server listed in servers.
concurrent_connections = 100
# If you comment out this setting, load will be simulated for as long
# as the program runs. This is the number of times each load definition will be 
# run for writes. Queries happen on intervals for as long as the benchmark is running.
runs_per_load_definition = 10000

[[servers]]
connection_string = "localhost:8086"
is_secure = false
skip_verify = false
timeout = "10s"

# Load definitions describe the reads and writes that you want to simulate.
# The connections take writes from the load definitions as they're sent out.
# You can define multiple load definitions that will run in parallel.
[[load_definitions]]
# The name is used only for reporting purposes. Response times and point counts
# will be reported to:
# <name>.ok
# <name>.fail
# The fail time series will contain the error that was returned by the go library.
name = "write_10_series"
# For load definitions that will be writing many hundreds of thousands of points
# it may be more desirable to do some sampling from successful writes. If you 
# comment out the following line you'll get the repsonse time from every 100th
# write reqeust reported. All failures will still be reported.
# report_sampling_interval = 100

# If you're doing sampling you'll want to report some percentiles and the max response
# times in some window of time. In addition to the sampled report stream you can get
# percentiles using these two variables. They would produce these series in the 
# reporting dataabase:
# <name>.percentiles.90, <name>.percentiles.95, <name>.percentiles.99, <name>.percentiles.max
# percentiles = [90.0, 95.0, 99.0]
# percentile_time_interval = "10s"

# series name will start with this string
base_series_name = "some_series"

# the number of time series to create
series_count = 100

  # write settings define write load that will be run against the server
  [load_definitions.write_settings]
  # the number of series that will be in each POST request. if less than series
  # count then multiple posts will be made on each 
  batch_series_size = 10

  # the number of points per series that will be in each POST
  batch_points_size = 100

  # time to wait between each POST request on each connection.
  # can be like 10u (microseconds), 10ms, 10s
  delay_between_posts = "0s"

  # each series in this load definition can have multiple int, string, bool, or float columns.
  [[load_definitions.int_columns]]
  name = "value"
  # int values will be random in [0, max_value)
  max_value = 10

  # bool values will be randomly true or false
  [[load_definitions.bool_columns]]
  name = "some_bool"

  # floats will be a random values in [0, 1.0)
  [[load_definitions.float_columns]]
  name = "some_other_val"

  [[load_definitions.string_columns]]
  name = "type"
  # values are potential column values. Each point will have one of these randomly selected
  values = ["click", "open", "view", "delete"]
  # if you want to simulate random string values (i.e. many possible unique strings)
  # leave 'values' out and use this argument. Random strings of that lenght will be generated
  # random_length = 50

  # multiple queries can be run against the server at the same time
  # the writes are happening.
  [[load_definitions.queries]]
  # name is for reporting purposes only. every successful query response time will get reported to:
  # <load definition name>.query.<name>.ok
  # <load definition name>.query.<name>.fail
  name = "count"

  # Using quer_start and query_end will execute against each time series this load definition
  # creates. The series names will be filled in automatically. This example will execute this
  # query against all 10 series every 10 seconds.
  query_start = "select count(value) from "
  query_end = " where time > now() - 30s"

  # queries run periodically based on this setting
  perform_every = "10s"

  [[load_definitions.queries]]
  name = "select_last_point_from_all"
  # The other option is to use full_query. This is useful for specific queries you want
  # to test out like selecting against a regex. This query won't be modified and it will
  # be executed once for every interval (regardless of how many series the load definition creates)
  full_query = "select * from /.*/ limit 1"
  perform_every = "5s"