Guide

Getting Started

Learn how to deploy LogStrata's controller daemon and fluentd collection agents in minutes.

System Prerequisites

Before installing LogStrata, verify that your target environment contains the following tools:

Kubernetes cluster version v1.24+
Helm package manager installed locally
Aggregated Elasticsearch database context or open port access
Write access to patch Deployment scale limits

1. Install via Helm

Add the official Helm repository context and pull down the LogStrata platform package:

# Register registry
$ helm repo add logstrata https://helm.logstrata.io
$ helm repo update

# Install charts into custom namespace
$ helm install logstrata logstrata/logstrata \
--namespace logstrata-system \
--create-namespace

2. Verify Ingestion

Ensure that the collection agent DaemonSet and Controller components are running correctly:

$ kubectl get pods -n logstrata-system

Expected output:

NAME                                   READY   STATUS    RESTARTS   AGE
logstrata-controller-5fc6d5b78-xyz12   1/1     Running   0          2m
fluentd-logging-agent-j4k2s            1/1     Running   0          2m
fluentd-logging-agent-l8f9d            1/1     Running   0          2m

3. Apply Your First Scaling Policy

Create a simple policy configuration to automatically scale your application on high request latency. Save the file as slsa-policy.yaml:

apiVersion: core.logstrata.io/v1alpha1
kind: LogAutoscalerPolicy
metadata:
  name: frontend-latency-scaler
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: main-frontend
  metricSources:
    - type: ElasticSearchQuery
      elasticsearch:
        query: "status:200 AND path:/home"
        timeWindow: "30s"
        trigger:
          metricName: throughput_rps
          threshold: 300.0
          scaleFactor: 1.5
  cooldownPeriod: "45s"
  minReplicas: 3
  maxReplicas: 15

Apply it directly to your cluster context:

$ kubectl apply -f slsa-policy.yaml

Next Steps

Read about how the internal architecture manages metric querying, or explore the configuration variables.