Skip to main content

Install Arcanna.ai

Arcanna can be installed using one of the following methods:

  1. AWS Marketplace (Recommended for quick deployment on AWS)
  2. Google Marketplace (Recommended for quick deployment on GCP)
  3. Arcanna Installer (Recommended for on-premise and private cloud installations)
  4. Kubernetes Installer (Recommended for managed kubernetes cluster installations)

1. Installing Arcanna from AWS Marketplace​

If you're deploying Arcanna on AWS, you can use our AWS Marketplace listing for a streamlined installation.

This method allows for faster deployment and seamless integration with AWS infrastructure. After launching the instance, follow the setup instructions provided in the Usage Instructions of the product or check our AWS Marketplace Installation Tutorial.

For any issues, refer to our support page or contact us directly.

2. Installing Arcanna from Google Marketplace​

If you're deploying Arcanna on GCP, you can use our Google Marketplace listing for a streamlined installation.

This method allows for faster deployment and seamless integration with GCP infrastructure. After launching the instance, follow the setup instructions provided in the Usage Instructions of the product or check our Google Marketplace Installation Tutorial.

For any issues, refer to our support page or contact us directly.

3. Installing Arcanna using the Arcanna Installer​

At the moment, the installation kit is available only via support. You can contact us at our support page.

System requirements​

Operating system:

  • Ubuntu 20 (focal), 22 (jammy), 24 (noble)
  • Debian 10 (buster), 11 (bullseye), 12 (bookworm)
  • RedHat/CentOS/Oracle Linux 8.0 or 9.0

Recommendation:

  • Swap partition should be disabled on the provisioned VM otherwise it will degrade the overall performance of Arcanna.ai.

Dependencies:

  • Docker should be installed, otherwise the Arcanna installer will install Docker for you.

Virtualization:

  • AVX or other specific flags should be enabled for the VM. While VMware has this configuration enabled by default, Proxmox requires manual configuration by setting up the CPU type to host from VM advanced settings.

Hardware requirements​

Minimum specifications:​

The bare minimum needed for the Arcanna to run. Installation is not allowed if the system doesn't meet these requirements.

8CPU32GB RAM70GB SSD
danger

We do not recommend running Arcanna with only the minimum specifications!

These are intended only for development and basic testing to verify that Arcanna can start.

Several functionalities will be limited or non-functional, and performance will be significantly impacted in a production environment.

Acceptable specifications:​

12CPU48GB RAM200GB SSD
warning

Arcanna will run with these specifications, but there can be potential performance issues, especially as the number of Pipelines and the system load increases.

For a stable production environment, we strongly recommend using the specifications below.

To ensure optimal performance and user experience, please configure your system as mentioned below.

16CPU64GB RAM1TB SSD
info

The required VM size depends on the number of alerts Arcanna needs to process, so it may vary from case to case.

Type of deployments available​

  • Standalone - single VM deployment
  • Distributed - deployment on top of Kubernetes

Arcanna can also be installed in offline environments without requiring internet access.

Upgrade Arcanna.ai​

To upgrade, simply run the installer for the newer version. Be sure to check with the support team for any potential version incompatibilities. Typically, all necessary migrations are performed automatically during the upgrade process.

However, we still recommend checking the Known Issues page before upgrading.

Uninstall Arcanna.ai​

To uninstall Arcanna, go to the arcanna-install path. This is usually ~/arcanna-install, or /root/arcanna-install if using root at install. Run the script bash ./scripts/uninstall_arcanna.sh.

4. Installing Arcanna using the Kubernetes Installer​

If you already operate a Kubernetes cluster — managed (EKS, GKE, AKS, OpenShift on cloud) or on-premise (RKE2, kubeadm, vanilla K8s) — you can install Arcanna directly using the arcanna-helm Helm charts and bundled Makefile.

This method gives you full control over Kubernetes-native concerns: namespacing, secret management, NodePort vs Ingress, storage classes, and GitOps integration via Argo CD or Flux.

For the complete step-by-step walkthrough — prerequisites, environment configuration, initial deploy, upgrades, secrets, networking, and GitOps, ask Arcanna team for the more comprehensive documentation.

System requirements​

A working Kubernetes cluster reachable via kubectl. Arcanna does not install Kubernetes itself.

Supported / tested distributions:

  • Cloud-managed: Amazon EKS 1.28+, Google GKE 1.28+, Azure AKS 1.28+
  • On-premise: RKE2, kubeadm-based clusters, vanilla Kubernetes 1.28+
  • OpenShift: ROSA, ARO, OpenShift on GCP, OCP 4.12+

Required cluster operators (installed once, before Arcanna):

  • ECK — Elastic Cloud on Kubernetes (manages Elasticsearch + Kibana)
  • CFK — Confluent for Kubernetes (manages Kafka + KRaft)

Tools on the operator workstation:

  • kubectl 1.28+, helm 3.14+, make (GNU), yq (mikefarah v4+), jq, openssl, bash, git

Hardware requirements​

The cluster as a whole, across all worker nodes:

Minimum specifications:​

The bare minimum needed for Arcanna to start. The deploy will fail if the cluster can't schedule all pods.

12 vCPU48GB RAM200GB SSD (cluster-wide)
danger

We do not recommend running Arcanna on the minimum cluster size.

These specs only verify that Arcanna can start. Workers, hypervisor, and Elasticsearch will be CPU/memory-throttled, and the pipeline throughput will be severely limited.

For high-throughput SOC environments processing 10K+ alerts/day with multiple active pipelines.

5+ nodes × 16 vCPU5+ nodes × 64GB RAM1TB+ SSD per node
info

The required cluster size depends on the number of alerts Arcanna processes and how many concurrent pipelines you run.

Storage requirements​

Arcanna's stateful workloads (Elasticsearch, Kafka, Postgres, Redis) all request ReadWriteOnce block storage on an SSD-backed StorageClass. Arcanna's model should sit on a ReadWriteMany block storage.

PlatformRecommended classProvided by
EKSgp3 (you create it on 1.30+)EBS CSI driver
GKEpremium-rwo (pre-installed)PD CSI driver
AKSmanaged-csi-premium (pre-installed)Azure Disk CSI driver
Bare-metal / RKE2sc-default via OpenEBS, Longhorn, Rook+Ceph, etc.You install a CSI driver

For full per-platform setup commands and the OpenEBS LVM example, ask Arcanna Team.

Type of deployments available​

  • Single namespace all of Arcanna in one K8s namespace (default arcanna); the recommended pattern
  • Multi-environment multiple namespaces on the same cluster (e.g. arcanna-stage, arcanna-e2e), driven by per-env value files under envs/<env>/
  • Multi-cluster (GitOps) Argo CD or Flux managing Arcanna across several clusters from a single deploy repo. Arcanna images are hosted on Google Container Registry (gcr.io/siscale-aiops/). The cluster needs a gcr-pull-secret to pull them contact support for the service account JSON key.

Quick start​

If your cluster meets the prerequisites and ECK + CFK are installed:

git clone https://github.com/siscale/arcanna-helm.git
cd arcanna-helm
git checkout v1.83.0

cp .env.example .env
# Edit .env at minimum set GCR_JSON_KEY_FILE

make deploy-all \
  ENV=baremetal-stage \
  NAMESPACE=arcanna-stage \
  TAG=v1.83.0

The Makefile runs a 10-phase ordered pipeline: infra → secrets → main-config → migration → core services → workers → monitoring → frontend → admin user → healthcheck. Total time is 15-30 minutes on a cold cluster.

The admin user's password is printed once at the end of the deploy capture it before closing the terminal.

Upgrade Arcanna.ai​

To upgrade to a newer release, pin the chart version and re-run the upgrade target:

cd arcanna-helm
git fetch && git checkout tags/v1.83.0

make upgrade-all \
  ENV=baremetal-stage \
  NAMESPACE=arcanna-stage \
  TAG=v1.84.0

upgrade-all skips the infra phase and only updates app services. ES index migrations run idempotently. For per-service upgrades, rollback, and password rotation.

We recommend checking the Known Issues page before upgrading.

Uninstall Arcanna.ai​

The Helm-based install offers four destroy levels, increasing in scope:

make destroy-app        ENV=<env> NAMESPACE=<ns>   # uninstall releases, keep PVCs+secrets
make destroy-data       ENV=<env> NAMESPACE=<ns>   # wipe PVCs and installer secrets
make destroy-all        ENV=<env> NAMESPACE=<ns>   # everything inside the namespace
make destroy-namespace  NAMESPACE=<ns>             # remove the namespace itself