Skip to content

Latest commit

 

History

History
96 lines (76 loc) · 4.05 KB

28-Dec'23.md

File metadata and controls

96 lines (76 loc) · 4.05 KB

Harness Performance Test Report - 28'DEC 2023

This document details information about following :

  1. Test Environment and resource configuration
  2. Test scenario and results

Environment

  • GKE (Kubernetes Version) : 1.25.x

Database

  • Mongo Atlas M60

Redis

  • GCP Memory Store (5GB)

Harness Services

Helm chart : https://github.com/harness/helm-charts/releases/tag/harness-0.11.2

Service Name Replicas CPU (per replica) Memory (per replica) Version
access-control 5 1 5 harness-0.11.2
ci-manager 6 3 6 harness-0.11.2
pipeline-service 10 4 12 harness-0.11.2
manager 6 3 12 harness-0.11.2
log-service 3 3 12 harness-0.11.2
ng-manager 4 3 8 harness-0.11.2
scm 2 0.5 1 harness-0.11.2
gateway 2 2 6 harness-0.11.2
default-backend 1 0.1 0.2 harness-0.11.2
nginx-ingress-controller 2 5 10 harness-0.11.2
change-data-capture 1 4 5 harness-0.11.2
next-gen-ui 2 0.5 0.5 harness-0.11.2
ng-auth-ui 2 0.1 0.1 harness-0.11.2
platform-service 2 0.5 3 harness-0.11.2
template-service 2 1 8 harness-0.11.2
ti-service 1 3 6 harness-0.11.2
sto-core 1 0.5 0.75 harness-0.11.2
sto-manager 1 3 6 harness-0.11.2
gitops 1 2 2 harness-0.11.2
ui 1 0.5 0.5 harness-0.11.2
policy-mgmt 1 0.5 0.5 harness-0.11.2
timescaledb 2 1 2 harness-0.11.2

Override file : https://github.com/harness/helm-charts/blob/main/src/harness/override-perf-ci-cd-ff.yaml

Manager config : update LOG_STREAMING_SERVICE_EXTERNAL_URL = <smp host url>/log-service/

Test Scenarios

> 1800 concurrent CI Executions [INLINE]

Each CI pipeline would

  • initialise a k8s pod and git clone repo
  • run 5 parallel steps (100 sec sleep)
  • run template with 2 parallel steps (140sec sleep)

Projects : 1
Pipelines : 1800
Stages per pipeline : 1
Delegates : 15 (1cpu/2gi) trigger : webhook

Result : PASS
Total Execution Time: 7.2min

> 1500 concurrent CI Executions [GitX]

Each CI pipeline would

  • initialise a k8s pod and git clone repo
  • run 5 parallel steps (360 sec sleep) and echo statements

Projects : 1
Pipelines : 1500
Stages per pipeline : 1
Delegates : 15 (1cpu/2gi) trigger : webhook

Result : PASS
Total Execution Time: 10.3min

> 1000 concurrent CD Executions [INLINE]

Each CD pipeline would

  • fetch docker artifact from AWS ECR repo
  • run following steps in order:
    • Canary deploy
    • Canary delete
    • Rolling deploy
    • K8s Delete

Projects : 1
Pipelines : 1000
Stages per pipeline : 1
Delegates : 27 (1cpu/4gi)

Result : PASS
Total Execution Time: 4.5min