performance-docs/doc/source/test_results/fuel_ccp_upgrade/index.rst

Fuel Containerized Control Plane performance report during upgrade

Abstract

This document includes Fuel CCP control plane performance test results for various environments (from 50 to 350 nodes). All tests have been performed regarding fuel_ccp_upgrade_test_plan

Environment description

For Kubernetes pre-deployment Kargo tool was used. More information about fuel-ccp and how it can be installed can be found in official documentation.

Hardware configuration of each server

All servers have configuration describing in table below

Description of servers hardware

server	vendor,model	HP,DL380 Gen9
CPU	vendor,model ----------------+ processor_count ----------------+ core_count ----------------+ frequency_MHz	Intel,E5-2680 v3 -------------------------------+ 2 -------------------------------+ 12 -------------------------------+ 2500
RAM	vendor,model ----------------+ amount_MB	HP,752369-081 -------------------------------+ 262144
NETWORK	interface_name ----------------+ vendor,model ----------------+ bandwidth	p1p1 -------------------------------+ Intel,X710 Dual Port -------------------------------+ 10G
STORAGE	dev_name ----------------+ vendor,model ----------------+ SSD/HDD ----------------+ size	/dev/sda -------------------------------+ | raid10 - HP P840 | 12 disks EH0600JEDHE -------------------------------+ HDD -------------------------------+ 3,6TB

Network configuration of each server

All servers have same network configuration:

Fuel CCP configuration

CCP main configuration ccp.yaml <configs/ccp.yaml>

CCP Cluster configuration configs.yaml <configs/configs.yaml>

CCP Cluster topology topology.yaml <configs/topology.yaml>

CCP repositories configuration repos.yaml <configs/repos.yaml>

CCP versions configuration for Mitaka versions-mitaka.yaml <configs/versions-mitaka.yaml>

CCP versions configuration for upgrade from Mitaka to Newton versions-newton.yaml <configs/versions-newton.yaml>

Rally configuration and scenario

Rally deployment rally_deployment.json <configs/rally_deployment.json>

Rally scenario create_and_delete.yaml <configs/create_and_delete.yaml>

Upgrade and tests procedure

• Install fuel-ccp
• Rename ccp.yaml to /root/.ccp.ymal
• Copy all yaml files to /root/
• Rename versions-mitaka.yaml to versions.yaml
• Execute ccp deploy and wait ok status for all services
• Create rally deployment with rally_deployment.json
• Create role for rally openstack --insecure role create member
• Start rally tests rally task start --task-args 'gre_enabled: true' ./create_and_delete.yaml
• Rename versions-newton.yaml to versions.yaml
• Execute ccp deploy and wait ok status for all services
• Wait tests results from rally

Test results

Test Case 1: Boot and delete server

The following set of results is dedicated to investigate how Nova installed against Kubernetes cluster via fuel-ccp tool is behaving during update from Mitaka to Newton. Upgrade started on 100 iteration and finished on the 156 iteration.

149 nodes OpenStack cluster, concurrency 5, 1530 iterations ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

NovaServers.boot_and_delete_server scenario in upgrade_newton_to_mitaka.html <reports/upgrade_newton_to_mitaka.html>

Operation	Median (sec)	90%ile (sec)	95%ile (sec)	Max (sec)	Min (sec)
boot_server	11.879	16.753	18.708	48.84	7.563
delete_server	2.593	4.754	4.898	6.829	2.313

The control plane downtime of keystone upgrades will be avoided in the feature versions of fuel-ccp