How to send JMeter live metrics to InfluxDB

Live load-test telemetry matters when a run is too long or too visible to wait for the final .jtl file. Apache JMeter can stream sampler metrics to InfluxDB through a Backend Listener, giving QA engineers and SREs a running view of throughput, latency, and errors while the test is still active.

The built-in InfluxdbBackendListenerClient sends line-protocol metrics over HTTP. For InfluxDB 2.x, the listener URL points at the write endpoint and includes the organization and bucket query parameters, while the token travels as a listener parameter instead of being embedded in the URL.

Use a bucket dedicated to load-test telemetry and a write-scoped token for the JMeter run. Keep the token out of source-controlled .jmx files by passing it at runtime with a -J property, then verify the bucket with a separate read-capable token or dashboard account.

1. Prepare the InfluxDB endpoint and token scope for the test run.

   InfluxDB write URL:
   http://influxdb:8086/api/v2/write
   ?org=load-team&bucket=jmeter
   
   JMeter write token:
   write access to the jmeter bucket
   
   Verification token:
   read access to the jmeter bucket
   
   Application tag:
   checkout-api

   Replace the host, organization, bucket, and application name with values from the environment that owns the load-test dashboard.

2. Add a Backend Listener to the test plan with the InfluxDB client and HTTP sender settings.

   Backend Listener implementation:
   InfluxdbBackendListenerClient
   
   influxdbMetricsSender=HttpMetricsSender
   influxdbUrl=${__P(influxdbUrl)}
   influxdbToken=${__P(influxdbToken)}
   application=checkout-api
   measurement=jmeter
   summaryOnly=false
   samplersRegex=checkout-.*
   percentiles=90;95;99
   testTitle=checkout smoke
   eventTags=env=staging

   The GUI may display fully qualified class names for the implementation and sender fields. Choose the built-in InfluxDB client and the JMeter HTTP metrics sender.

3. Store the write and read tokens in the shell that will start JMeter and verify the bucket.

   $ export INFLUXDB_WRITE_TOKEN='paste-write-token-here'
   $ export INFLUXDB_READ_TOKEN='paste-read-token-here'

   Do not commit tokens into the .jmx file, CI configuration, or shared result artifacts. Pass them from a secret manager, protected CI variable, or local shell environment.

4. Run the test plan in non-GUI mode and pass the listener values as JMeter properties.

   $ jmeter -n \
     -t checkout-load-test.jmx \
     -l checkout-load-test.jtl \
     -JinfluxdbUrl='http://influxdb:8086/api/v2/write?org=load-team&bucket=jmeter' \
     -JinfluxdbToken="$INFLUXDB_WRITE_TOKEN"
   Creating summariser <summary>
   Created the tree successfully using checkout-load-test.jmx
   Starting standalone test @ 2026 Jun 29 21:28:26 GMT
   summary = 6 in 00:00:00 = 78.9/s Avg: 2 Min: 1 Max: 12 Err: 0 (0.00%)
   Tidying up ...

   The summary confirms that JMeter ran the plan and finished without failed samples. The Backend Listener flushes its final metrics at the end of the run.

5. Query InfluxDB for the sampler count rows written by the Backend Listener.

   $ influx query --raw \
     --org load-team \
     --token "$INFLUXDB_READ_TOKEN" '
   from(bucket:"jmeter")
     |> range(start: -30m)
     |> filter(fn: (r) => r.application == "checkout-api")
     |> filter(fn: (r) => r.transaction == "checkout-home")
     |> filter(fn: (r) => r._field == "count")
   '
   #group,false,false,false,true,true,true,true
   #datatype,string,long,double,string,string,string,string
   #default,_result,,,,,,
   ,result,table,_value,_field,application,statut,transaction
   ,,0,6,count,checkout-api,all,checkout-home
   ,,1,6,count,checkout-api,ok,checkout-home

   Rows with application=checkout-api and transaction=checkout-home prove that the Backend Listener wrote metrics for the running test plan, not only that JMeter produced a local result file.

6. Clear the temporary shell tokens after the run is verified.

   $ unset INFLUXDB_WRITE_TOKEN INFLUXDB_READ_TOKEN