https://github.com/rsevilla87/k8s-dnsperf

Kubernetes DNS loader
https://github.com/rsevilla87/k8s-dnsperf

client-go dns dnsperf golang kubernetes openshift performance

Last synced: about 1 month ago
JSON representation

Kubernetes DNS loader

• Host: GitHub
• URL: https://github.com/rsevilla87/k8s-dnsperf
• Owner: rsevilla87
• License: apache-2.0
• Created: 2024-09-25T14:43:17.000Z (4 months ago)
• Default Branch: main
• Last Pushed: 2024-10-26T23:13:45.000Z (2 months ago)
• Last Synced: 2024-10-28T02:49:48.119Z (2 months ago)
• Topics: client-go, dns, dnsperf, golang, kubernetes, openshift, performance
• Language: Go
• Homepage:
• Size: 43.9 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# k8s-dnsperf

Kubernetes DNS loader

## Running

k8s-dnsperf is a tool that helps to run [dnsperf](https://github.com/DNS-OARC/dnsperf) in a kubernetes cluster. It provides multiple features.

A regular k8s-dnsperf execution consists of several stages:

- Deploy benchmark assets

  - k8s-dnsperf namespace

  - k8s-dnsperf DaemonSet, initially in idle status

  - ClusterIP Services, each DNS record is generated from a k8s service, first record will always be `kubernetes.default.svc.cluster.local`.

  - ConfigMap holding the previous records

- Run dnsperf benchmark using the configured flags in parallel against the target DNS server, by default `172.30.0.10`

- Parse dnsperf output

- Index results. (Indexing is optional)

- Destroy assets

```shell

$ ./bin/k8s-dnsperf -duration 1m --es-server ${ES_SERVER}

2024-10-08 13:00:27 INF cmd/k8s-dnsperf.go:41 > Starting k8s-dnsperf main@e056503b9e2c6a4db7e7b02e4cf2d83220b668c9

2024-10-08 13:00:27 INF cmd/k8s-dnsperf.go:43 > Creating indexer 💾

2024-10-08 13:00:29 INF pkg/infra/infrastructure.go:56 > Creating benchmark assets 🚧

2024-10-08 13:00:30 INF pkg/infra/infrastructure.go:106 > Waiting for DaemonSet k8s-dnsperf/k8s-dnsperf pods to be running

2024-10-08 13:00:36 INF pkg/benchmark/dnsperf.go:40 > Running benchmark 🚀

...

```

## Prometheus metrics

Several prometheus metrics and expressions can be used to monitor CoreDNS performance, some of the most interesting ones are described below:

- `sum(irate(container_memory_rss{name=~"k8s_dns_dns.+",namespace=~"openshift-dns"}[2m])) * 100`: Cluster CoreDNS aggregated CPU usage (100% means 1 core)

- `irate(container_memory_rss{name=~"k8s_dns_dns.+",namespace=~"openshift-dns"}[2m])`: Per instance CoreDNS CPU usage

- `sum(irate(coredns_dns_request_count_total{}[2m])) by (job)`: Aggregated CoreDNS queries per second

- `irate(coredns_dns_request_count_total{}[2m])`: Per instance CoreDNS queries per second

- `container_memory_rss{name=~"k8s_dns_dns.+",namespace=~"openshift-dns"}`: Per instance CoreDNS RSS memory in bytes

- `sum(container_memory_rss{name=~"k8s_dns_dns.+",namespace=~"openshift-dns"})`: Aggregated CoreDNS RSS memory in bytes

- `histogram_quantile(0.99, rate(coredns_dns_request_duration_seconds_bucket[2m]))`: P99 request latency in seconds

- `rate(coredns_cache_hits_total[2m])`: Cache hits rate

- `rate(coredns_cache_misses_total[2m])`: Cache miss rate

## Building

Requires go 1.22 or greater and make.