Skip to content

Latest commit

 

History

History
133 lines (111 loc) · 4.79 KB

README.md

File metadata and controls

133 lines (111 loc) · 4.79 KB
Raw

If you have arrived from the community blog-post please follow this repo's guide instead of the blog's, because there were some updates(p8s, ingress, elb) and the blog is not updated yet.

Loki-stack

Loki, Promtail, and Grafana are designed for efficient log aggregation and visualization. Loki, a horizontally scalable log storage system, seamlessly integrates with Promtail, a lightweight log shipper, while Grafana provides a user-friendly interface for exploring and analyzing log data, making the Loki Stack a robust solution for centralized logging and monitoring.
In this stack we will use an encrypted OBS bucket for Loki backend.

Image

Grafana

  • deployed as a statefulset with 1 replicas
  • uses EVS PVC for persistance
  • community dashboards out of the box

Loki

  • deployed in monolithic mode with 2 statefulset replicas
  • uses EVS PVC for persistance and caching
  • uses encrypted OBS to store indexes and log chunks
  • supports lifecycle policies via compactor and limits_config

Promtail

  • deployed as a deamonset
  • forwards pod and node logs to loki-gateway

Prometheus

  • deployed as statefulset with one replicas
  • kube-state-metrics and node-exporter included

Ingress ELB for Grafana

  • self-signed TLS certificate
  • Shared ELB and EIP managed by CCE ingress controller

OBS

  • encrypted OBS bucket for loki to store indexes and chunks
  • supports lifecycle policy via index_expiration terraform variables

RBAC

  • OTC user with the minimum amount of roles needed to access the encrypted OBS bucket

Dependencies

Infrastructure

  • Required:
    • CCE cluster with autoscaling
    • Internet access from VPC
    • Domain maintained in your OTC Tenant, and used for your Grafana URL/FQDN

Packages

Packages Dockerfile alternative

You need these packages installed. If you don't want to install them on your local machine you can use the Dockerfile to build an environment for the deployment:

docker build --no-cache=true --build-arg BUILD_DATE=$(date -u +'%Y-%m-%dT%H:%M:%SZ') --tag loki-deployer:latest .
docker run -it -v$(pwd):/deploy -v$HOME/.docker/config.json:/root/.docker/config.json -v$HOME/.kube/config:/root/.kube/config loki-deployer:latest

How to deploy the stack

Deploy cloud_services

  • RBAC for loki to write OBS
  • Encrypted OBS bucket
  • KMS keys
  1. Set up environment variables for terraform in .envrc
# update .envrc first
source .envrc && source get_token.sh
export GRAFANA_FQDN=[MY_GRAFANA_HOSTNAME]
  1. Set up OBS backend for terraform
tfenv install
terraform  -chdir=./cloud_services/tf_backend init && terraform -chdir=./cloud_services/tf_backend apply
  1. Deploy cloud_services
#Set the Loki bucket name and file retention time(OBS lifecycle for indexes) for cloud_services
echo "s3_chunks = \"MY_BUCKET_NAME\"" > cloud_services/terraform.tfvars
echo "index_expiration = 100" >> cloud_services/terraform.tfvars
#Set the VPC Subnet Name in which the ELB should be created
export TF_VAR_subnet_name='MY_SUBNET_NAME'
#Set the domain name where you need to add an A record
export TF_VAR_GRAFANA_DOMAIN=[MY_GRAFANA_DOMAIN]
export TF_VAR_GRAFANA_FQDN=$GRAFANA_FQDN
terraform -chdir=./cloud_services init && terraform -chdir=./cloud_services apply

Deploy on CCE

  1. Create kubernetes namespace, and set default storageclass to csi-disk-topology
kubectl create ns logging
kubectl patch storageclass csi-disk-topology -p '{"metadata": {"annotations":{"storageclass.kubernetes.io/is-default-class":"true"}}}'
  1. Set up environment variables for Loki deployment
source get_loki_creds.sh
  1. Generate a self-signed certificate for Grafana
./gen_self_signed.sh $TF_VAR_GRAFANA_FQDN
  1. Deploy the stack
helmfile sync
  1. Update your DNS record with the EIP of the ELB

Useful stuff

  • Get grafana admin password
kubectl get secret --namespace logging grafana -o jsonpath="{.data.admin-password}" | base64 --decode ; echo
  • Port-forward if you are not using ingress
kubectl port-forward svc/grafana 8080:80 -n logging

Uninstall the stack

helmfile destroy && kubectl delete pvc --all -nlogging
terraform -chdir=./cloud_services destroy
terraform -chdir=./cloud_services/tf_backend destroy
#KMS keys will be removed after 1 week grace period