How to Perform Load Testing against NebulaGraph with K6

Load Testing of K6 in NebulaGraph

Why Load Testing Matters in NebulaGraph?

The load testing for the database needs to be conducted usually so that the impact on the system can be monitored in different scenarios, such as query language rule optimization, storage engine parameter adjustment, etc.

The operating system in this article is the x86 CentOS 7.8.

The hosts where NebulaGraph is deployed are configured with 4C 16G memory, SSD disk, and 10G network.

Tools Needed for the Load Testing

  • nebula-ansible deploys NebulaGraph services.
  • nebula-importer imports data into NebulaGraph clusters.
  • k6-plugin is a K6 extension that is used to perform load testing against the NebulaGraph cluster. The extension integrates with the nebula-go client to send requests during the testing.
  • nebula-bench generates the LDBC dataset and then imports it into NebulaGraph.
  • ldbc_snb_datagen_hadoop is a LDBC data generator.

Load Testing Process Overview

The load testing conducted in this article uses the LDBC dataset generated by ldbc_snb_datagen. The testing process is as follows.

Load Testing of K6 in NebulaGraph

To deploy the topology, use one host as the load testing runner, and use three hosts to form a NebulaGraph cluster.

Load Testing of K6 in NebulaGraph

To make monitoring easier, the load testing runner also deploys:

  • Prometheus
  • Influxdb
  • Grafana
  • node-exporter

The hosts where NebulaGraph is installed also deploy:

  • node-exporter
  • process-exporter

Load Testing Steps

Use nebula-ansible to deploy NebulaGraph

  1. Set up SSH login without passwords a. Log in,,, and respectively. Create a vesoft user and join in sudoer with NOPASSWD. b. Log in to set up SSH.

ssh-copy-id vesoft@
ssh-copy-id vesoft@
ssh-copy-id vesoft@
  1. Download nebula-ansible, install Ansible, and modify the Ansible configuration.
sudo yum install ansible -y
git clone https://github.com/vesoft-inc/nebula-ansible
cd nebula-ansible/

The following is an example of inventory.ini.

# GA or nightly
install_source_type = GA
nebula_version = 2.0.1
os_version = el7
arc = x86_64
pkg = rpm

packages_dir = {{ playbook_dir }}/packages
deploy_dir = /home/vesoft/nebula
data_dir = {{ deploy_dir }}/data

# ssh user
ansible_ssh_user = vesoft

force_download = False



  1. Install and deploy NebulaGraph.
ansible-playbook install.yml
ansible-playbook start.yml

Monitor hosts

Using docker-compose to deploy monitoring system is convenient. Docker and Docker-Compose need to be installed on the hosts first.

Log in

git clone https://github.com/vesoft-inc/nebula-bench.git

cd nebula-bench
cp -r third/promethues ~/.
cp -r third/exporter ~/.

cd ~/exporter/ && docker-compose up -d

cd ~/promethues
# Modify the exporter address of monitoring nodes
# vi prometheus.yml
docker-compose up -d

# Copy exporter to,, and, and then start docker-compose

Configure the Grafana data source and dashboard. For details, see https://github.com/vesoft-inc/nebula-bench/tree/master/third.

Generate the LDBC dataset

cd nebula-bench

sudo yum install -y git \
                    make \
                    file \
                    libev \
                    libev-devel \
                    gcc \
                    wget \
                    python3 \
                    python3-devel \
                    java-1.8.0-openjdk \

pip3 install --user -r requirements.txt

# Using `snb.interactive.1` parameter in ldbc_snb_datagen_hadoop. For more infor, see https://github.com/ldbc/ldbc_snb_datagen_hadoop/wiki/Configuration

python3 run.py data

# Date generated by mv

mv target/data/test_data/ ./sf1

Import data

cd nebula-bench
# Modify .evn
cp env .env
vi .env

The following is the example of .env

# Compile nebula-importer and K6

# Import data
python3 run.py nebula importer

During the import process, you can focus on the following network bandwidth and disk IO writing.

Import data

Import data

Execute the load testing

python3 run.py stress run

According to the code source in the file scenarios, the js file will be automatically rendered and K6 will be used to test all scenarios.

After the execution is over, the js file and the result will be saved in the output folder.

Among them, latency is the latency time returned by the server, and responseTime is the time from initiating execute to response by the client. The measurement unit is μs.

[vesoft@qa-60 nebula-bench]$ more output/result_Go1Step.json
    "metrics": {
        "data_sent": {
            "count": 0,
            "rate": 0
        "checks": {
            "passes": 1667632,
            "fails": 0,
            "value": 1
        "data_received": {
            "count": 0,
            "rate": 0
        "iteration_duration": {
            "min": 0.610039,
            "avg": 3.589942336582023,
            "med": 2.9560145,
            "max": 1004.232905,
            "p(90)": 6.351617299999998,
            "p(95)": 7.997563949999995,
            "p(99)": 12.121579809999997
        "latency": {
            "min": 308,
            "avg": 2266.528722763775,
            "med": 1867,
            "p(90)": 3980,
            "p(95)": 5060,
            "p(99)": 7999
        "responseTime": {
            "max": 94030,
            "p(90)": 6177,
            "p(95)": 7778,
            "p(99)": 11616,
            "min": 502,
            "avg": 3437.376111156418,
            "med": 2831
        "iterations": {
            "count": 1667632,
            "rate": 27331.94978169588
        "vus": {
            "max": 100,
            "value": 100,
            "min": 0
[vesoft@qa-60 nebula-bench]$ head -300 output/output_Go1Step.csv | grep -v USE
1628147822,GO 1 STEP FROM 4398046516514 OVER KNOWS,1217,1536,true,1,
1628147822,GO 1 STEP FROM 2199023262994 OVER KNOWS,1388,1829,true,94,
1628147822,GO 1 STEP FROM 1129 OVER KNOWS,1488,2875,true,14,
1628147822,GO 1 STEP FROM 6597069771578 OVER KNOWS,1139,1647,true,30,
1628147822,GO 1 STEP FROM 2199023261211 OVER KNOWS,1399,2096,true,6,
1628147822,GO 1 STEP FROM 2199023256684 OVER KNOWS,1377,2202,true,4,
1628147822,GO 1 STEP FROM 4398046515995 OVER KNOWS,1487,2017,true,39,
1628147822,GO 1 STEP FROM 10995116278700 OVER KNOWS,837,1381,true,3,
1628147822,GO 1 STEP FROM 933 OVER KNOWS,1130,3422,true,5,
1628147822,GO 1 STEP FROM 6597069771971 OVER KNOWS,1022,2292,true,60,
1628147822,GO 1 STEP FROM 10995116279952 OVER KNOWS,1221,1758,true,3,
1628147822,GO 1 STEP FROM 8796093031179 OVER KNOWS,1252,1811,true,13,
1628147822,GO 1 STEP FROM 10995116279792 OVER KNOWS,1115,1858,true,6,
1628147822,GO 1 STEP FROM 6597069777326 OVER KNOWS,1223,2016,true,4,
1628147822,GO 1 STEP FROM 8796093028089 OVER KNOWS,1361,2054,true,13,
1628147822,GO 1 STEP FROM 6597069777454 OVER KNOWS,1219,2116,true,2,
1628147822,GO 1 STEP FROM 13194139536109 OVER KNOWS,1027,1604,true,2,
1628147822,GO 1 STEP FROM 10027 OVER KNOWS,2212,3016,true,83,
1628147822,GO 1 STEP FROM 13194139544176 OVER KNOWS,855,1478,true,29,
1628147822,GO 1 STEP FROM 10995116280047 OVER KNOWS,1874,2211,true,12,
1628147822,GO 1 STEP FROM 15393162797860 OVER KNOWS,714,1684,true,5,
1628147822,GO 1 STEP FROM 6597069770517 OVER KNOWS,2295,3056,true,7,
1628147822,GO 1 STEP FROM 17592186050570 OVER KNOWS,768,1630,true,26,
1628147822,GO 1 STEP FROM 8853 OVER KNOWS,2773,3509,true,14,
1628147822,GO 1 STEP FROM 19791209307908 OVER KNOWS,1022,1556,true,6,
1628147822,GO 1 STEP FROM 13194139544258 OVER KNOWS,1542,2309,true,91,
1628147822,GO 1 STEP FROM 10995116285325 OVER KNOWS,1901,2556,true,0,
1628147822,GO 1 STEP FROM 6597069774931 OVER KNOWS,2040,3291,true,152,
1628147822,GO 1 STEP FROM 8796093025056 OVER KNOWS,2007,2728,true,29,
1628147822,GO 1 STEP FROM 21990232560726 OVER KNOWS,1639,2364,true,9,
1628147822,GO 1 STEP FROM 8796093030318 OVER KNOWS,2145,2851,true,6,
1628147822,GO 1 STEP FROM 21990232556027 OVER KNOWS,1784,2554,true,5,
1628147822,GO 1 STEP FROM 15393162796879 OVER KNOWS,2621,3184,true,71,
1628147822,GO 1 STEP FROM 17592186051113 OVER KNOWS,2052,2990,true,5,

It is also possible to execute the load testing in one scenario and continuously adjust the configuration parameters for comparison.

Concurrent reading
# Run Go2Step with 50 virtual users and 300 seconds of duration
python3 run.py stress run -scenario go.Go2Step -vu 50 -d 300
INFO[0302] 2021/08/06 03:55:27 [INFO] finish init the pool

     ✓ IsSucceed

     █ setup

     █ teardown

     checks...............: 100.00% ✓ 1559930     ✗ 0
     data_received........: 0 B     0 B/s
     data_sent............: 0 B     0 B/s
     iteration_duration...: min=687.47µs avg=9.6ms       med=8.04ms max=1.03s  p(90)=18.41ms p(95)=22.58ms p(99)=31.87ms
     iterations...........: 1559930 5181.432199/s
     latency..............: min=398      avg=6847.850345 med=5736   max=222542 p(90)=13046   p(95)=16217   p(99)=23448
     responseTime.........: min=603      avg=9460.857877 med=7904   max=226992 p(90)=18262   p(95)=22429   p(99)=31726.71
     vus..................: 50      min=0         max=50
     vus_max..............: 50      min=50        max=50

Every metric can be monitored at the same time.

Concurrent reading

checks is to verify whether the request is executed successfully. If the execution fails, the failed message will be saved in the CSV file.

awk -F ',' '{print $NF}' output/output_Go2Step.csv|sort |uniq -c
# Execute Go2Step with 200 virtual users and 300 seconds of duration
python3 run.py stress run -scenario go.Go2Step -vu 200 -d 300
INFO[0302] 2021/08/06 04:02:34 [INFO] finish init the pool

     ✓ IsSucceed

     █ setup

     █ teardown

     checks...............: 100.00% ✓ 1866850    ✗ 0
     data_received........: 0 B     0 B/s
     data_sent............: 0 B     0 B/s
     iteration_duration...: min=724.77µs avg=32.12ms      med=25.56ms max=1.03s  p(90)=63.07ms p(95)=84.52ms  p(99)=123.92ms
     iterations...........: 1866850 6200.23481/s
     latency..............: min=395      avg=25280.893558 med=20411   max=312781 p(90)=48673   p(95)=64758    p(99)=97993.53
     responseTime.........: min=627      avg=31970.234329 med=25400   max=340299 p(90)=62907   p(95)=84361.55 p(99)=123750
     vus..................: 200     min=0        max=200
     vus_max..............: 200     min=200      max=200

K6 metrics to be monitored with Grafana

The monitoring data of K6 in Grafana

Concurrent writing
# Execute insert with 200 virtual users and 300 seconds of duration. By default, batchSize is 100.

python3 run.py stress run -scenario go.Go2Step -vu 200 -d 300

The js file can be modified manually to adjust batchSize

sed -i 's/batchSize = 100/batchSize = 300/g' output/InsertPersonScenario.js

# Run K6 manually

scripts/k6 run output/InsertPersonScenario.js -u 400 -d 30s --summary-trend-stats "min,avg,med,max,p(90),p(95),p(99)" --summary-export output/result_InsertPersonScenario.json --out influxdb=

If the batchSize is 300 with 400 virtual users, an error will be returned.

INFO[0032] 2021/08/06 04:03:49 [INFO] finish init the pool

     ✗ IsSucceed
      ↳  96% — ✓ 31257 / ✗ 1103

     █ setup

     █ teardown

     checks...............: 96.59% ✓ 31257       ✗ 1103
     data_received........: 0 B    0 B/s
     data_sent............: 0 B    0 B/s
     iteration_duration...: min=12.56ms avg=360.11ms      med=319.12ms max=2.07s   p(90)=590.31ms p(95)=696.69ms p(99)=958.32ms
     iterations...........: 32360  1028.339207/s
     latency..............: min=4642    avg=206931.543016 med=206162   max=915671  p(90)=320397.4 p(95)=355798.7 p(99)=459521.39
     responseTime.........: min=6272    avg=250383.122188 med=239297.5 max=1497159 p(90)=384190.5 p(95)=443439.6 p(99)=631460.92
     vus..................: 400    min=0         max=400
     vus_max..............: 400    min=400       max=400
awk -F ',' '{print $NF}' output/output_InsertPersonScenario.csv|sort |uniq -c
   1103  error: E_CONSENSUS_ERROR(-16)."
      1 errorMsg 

If E_CONSENSUS_ERROR occurs, it should be that the appendlog buffer of raft is overflown when the concurrency is large, which can be solved by adjusting relevant parameters.


  • The load testing uses the LDBC dataset standard to ensure data uniform. Even when bigger data volume, say one billion vertices, is generated, the graph schema is the same.
  • K6 is more convenient than Jmeter for the load testing. For more details, please refer https://k6.io/docs/.
  • You can easily find the bottleneck of the system resources by simulating various scenarios or adjust parameters in NebulaGraph with the mentioned tools.

If you encounter any problems in the process of using NebulaGraph, please refer to NebulaGraph Database Manual to troubleshoot the problem. It records in detail the knowledge points and specific usage of the graph database and the graph database NebulaGraph.

Join our Slack channel if you want to discuss with the rest of the NebulaGraph community!