Since JMeter 2.13 you can get real-time results sent to a backend through the
Backend Listener
using potentially any backend (JDBC, JMS, Webservice, …)
by providing a class which implements
AbstractBackendListenerClient
.
JMeter ships with a GraphiteBackendListenerClient which allows you to send metrics to a Graphite Backend.
This feature provides:
-
Live results
-
Nice graphs for metrics
-
Ability to compare 2 or more load tests
-
Storing monitoring data as long as JMeter results in the same backend
-
…
In this document we will present the configuration setup to graph and historize the data in 2 different backends:
16.1 Metrics exposed
|
16.1.1 Thread/Virtual Users metrics
|
Thread metrics are the following:
-
test.minAT
-
Min active threads
-
test.maxAT
-
Max active threads
-
test.meanAT
-
Mean active threads
-
test.startedT
-
Started threads
-
test.endedT
-
Finished threads
|
|
16.1.2 Response times metrics
|
Response related metrics are the following:
-
.ok.count
-
Number of successful responses for sampler name
-
.h.count
-
Server hits per seconds, this metric cumulates Sample Result and Sub results (if using Transaction Controller, "Generate parent sampler" should be unchecked)
-
.ok.min
-
Min response time for successful responses of sampler name
-
.ok.max
-
Max response time for successful responses of sampler name
-
.ok.avg
-
Average response time for successful responses of sampler name.
-
.ok.pct
-
Percentile computed for successful responses of sampler name. There will be one metric for each calculated value.
-
.ko.count
-
Number of failed responses for sampler name
-
.ko.min
-
Min response time for failed responses of sampler name
-
.ko.max
-
Max response time for failed responses of sampler name
-
.ko.avg
-
Average response time for failed responses of sampler name.
-
.ko.pct
-
Percentile computed for failed responses of sampler name. There will be one metric for each calculated value.
-
.a.count
-
Number of responses for sampler name (sum of ok.count and ko.count)
-
.a.min
-
Min response time for responses of sampler name (min of ok.count and ko.count)
-
.a.max
-
Max response time for responses of sampler name (max of ok.count and ko.count)
-
.a.avg
-
Average response time for responses of sampler name (avg of ok.count and ko.count)
-
.a.pct
-
Percentile computed for responses of sampler name. There will be one metric for each calculated value. (calculated on the totals for OK and failed samples)
The default
percentiles
setting on the
Backend Listener
is "90;95;99",
i.e. the 3 percentiles 90%, 95% and 99%.
The
Graphite naming hierarchy
uses dot (".") to separate elements. This could be confused with decimal percentile values.
JMeter converts any such values, replacing dot (".") with underscore ("-").
For example, "
99.9
" becomes "
99_9
"
By default JMeter sends metrics for all samplers accumulated under the samplerName "
all
".
If the Backend Listener
samplersList
is configured, then JMeter also sends the metrics
for the matching sample names unless
summaryOnly=true
|
|
|
|
16.2 InfluxDB
|
InfluxDB is an open-source, distributed, time-series database that allows to
easily store metrics.
Installation and configuration is very easy, read this for more details
InfluxDB documentation
.
InfluxDB data can be easily viewed in a browser through either
Influga
or
Grafana
.
We will use Grafana in this case.
16.2.3 Grafana configuration
|
Installing grafana is just a matter of putting the unzipped bundle behind an Apache HTTP server.
Read
documentation
for more details.
Open
config.js
file and find
datasources
element, and edit it like this:
 |
 |
 |
 |
datasources: {
influxdb: {
type: 'influxdb',
url: "http://influx_db_host:8086/db/jmeter",
username: 'root',
password: 'root',
},
grafana: {
type: 'influxdb',
url: "http://influx_db_host:8086/db/grafana",
username: 'root',
password: 'root',
grafanaDB: true
},
},
|
 |
 |
 |
 |
Note that grafana has "
grafanaDB:true
". Also note that here we use
root
user for simplicity
It is better to dedicate a special user with restricted rights.
|

Grafana dashboard
|
|
|
|
|
|