Service 01 · Planning and commissioning

Design & setup: from load profile to high availability architecture

In brief

High availability cluster design defines how a cluster must be built so your application survives component failures. At managed-cluster.net, we derive it from your load profile, availability target and recovery targets (RTO/RPO): sizing, redundancy with quorum, network and storage. You receive it in writing with an architecture drawing and a fixed price; we test failover before handover.

Scope

High availability cluster design: ten building blocks

Design & setup is the first of four service areas in our managed infrastructure operations. Our HA architecture consulting covers ten building blocks, from the initial assessment to documentation. We set out the plan in a written design, which then serves as the blueprint for building and testing your cluster.

  • Assessment and requirements analysis: application, load profile, availability target and recovery targets, meaning RTO (the maximum tolerable time until service is restored) and RPO (the maximum tolerable data loss, measured as the time span before the failure).
  • Cluster sizing: we discuss the hardware of the cluster nodes (CPU, RAM, storage, network) individually and define it in the design proposal, with headroom for load peaks and growth.
  • Redundancy concept: an N+1 design, in which the remaining nodes carry the load of a failed node, a quorum as the majority of nodes that keeps the cluster able to make decisions, and protection against split brain, where parts of a cluster keep running independently. Background: redundancy at every layer.
  • Virtualization: setup and configuration of the virtualization layer, for example Proxmox VE with KVM.
  • Load balancing design: redundant load balancers with health checks.
  • Network architecture: VLAN separation, firewalls and redundant connectivity; see network architecture on AS47263.
  • Storage and replication: a storage and replication design matched to data volume, I/O profile and RPO target.
  • Migration plan: moving your application with a minimal, predictable switchover window.
  • Failover test before handover: before handover, we test the failure of a node, and you receive the log. During operations, regular failover tests with a written log follow.
  • Documentation: architecture, configuration and emergency plan, in writing. It belongs to you.
Diagram · Design From load profile to written design
Your requirements shape sizing, redundancy, network, storage and the migration plan; the written design records them with an architecture drawing, component list and fixed price.
Diagram as text
  1. Inputs: application, load profile, availability target, RTO/RPO, data volume and I/O profile, existing environment.
  2. Design decisions: sizing, N+1 with quorum, split-brain protection, network and load balancing, storage replication, migration plan.
  3. Written design: architecture drawing, component list, fixed price for build and operations.
  4. Then: build and failover test before handover.

Process

Process: four steps to a high availability cluster

We design and set up a high availability cluster in four steps: assessment, architecture with a fixed price, build with a failover test, and migration and handover. You receive an initial assessment within one business day. The design and proposal are usually ready within a few business days; build and testing typically take one to three weeks.

  1. Step 01

    Assessment

    We review your application, load data and existing infrastructure and agree on the availability target and recovery targets with you.

    Timeframe initial assessment within one business day

  2. Step 02

    Architecture & fixed price

    We define sizing, redundancy, network and storage and set them out in a design with a fixed price.

    Timeframe usually within a few business days

  3. Step 03

    Build & failover test

    We build the cluster, set up monitoring and backups and run a full failover test before your application goes live on it.

    Timeframe typically one to three weeks

  4. Step 04

    Migration & handover

    Your application moves according to the migration plan. Then we take over operations with 24/7 infrastructure monitoring and on-call duty.

    Timeframe you set the date

Replacing VMware? We build the cluster alongside your existing environment, which stays ready to run until a successful cutover with a rollback plan. Details: migration plan for VMware environments.

Deliverables

Deliverables: what you receive at the end

At the end of design and setup, you have three written deliverables: the design with an architecture drawing, component list and fixed price for build and operations, the log of the failover test run before handover, and documentation of the architecture, configuration and emergency plan. The documentation belongs to you.

Deliverables Written deliverables of design and setup
DeliverableContentsWhen
DesignArchitecture drawing, component list, fixed price for build and operationsafter step 02
Test logLog of the full failover testafter step 03, before go-live
DocumentationArchitecture, configuration and emergency planat handover in step 04

During operations, you also receive a post-incident report for every incident and a monthly operations report. You have a dedicated contact who knows your cluster.

Components

Proxmox cluster planning: components of an HA architecture

At managed-cluster.net, components of a high availability design include, for example, Proxmox VE with KVM, replicated storage, HAProxy and keepalived, and database replication for PostgreSQL or MySQL/MariaDB. Which of them a design uses depends on your use case; the selection appears in the component list.

Examples Components by layer and their role in the cluster
LayerComponentRole in the cluster
VirtualizationProxmox VE with KVMIf a node fails, the Proxmox HA Manager restarts the affected virtual machines on another node.
StorageCeph or ZFS replicationCeph: three replicas of every block on different nodes, in the Essentials and Business packages. ZFS replication: replicates at set intervals, not synchronously; we use it in individually designed, more compact setups. The interval determines how much data can be lost in a failure.
Load balancingHAProxy and keepalivedRedundant load balancers with health checks and a floating IP. The load balancer removes a failed node from rotation within seconds.
DatabasePostgreSQL or MySQL/MariaDBReplication between database instances
NetworkVLANs, firewalls, bonding/LACPSeparate networks for management, storage and public traffic
OperationsMonitoring and backupsSet up during the build. Backups sit on separate infrastructure, because an HA cluster is no substitute for backups (backup & disaster recovery).

Existing hardware

Integrating existing hardware and colocation

A high availability cluster from managed-cluster.net does not require hardware or a data center of your own: it runs on infrastructure we manage in German data centers. If you already have servers or colocation space, we assess case by case whether and how they can be integrated into your cluster.

Our primary site is Maincubes FRA01; NBG01 in Nuremberg is the target for backup copies outside the primary site and the second site for multi-site deployments. More on choosing between FRA01 and NBG01.

Existing clusters

Take over a Proxmox cluster rather than rebuild it

If you already run a Proxmox cluster, we review it in an audit and document where it deviates from the target design. We then connect it to our monitoring and take over operations, either as it is or with a step-by-step move to the target architecture.

Have an existing Proxmox cluster managed

Preparation

Preparation: what to clarify before the design

For the assessment, we need information about your application, load profile, availability and recovery targets, and your existing environment. You can clarify the following seven points in-house in advance, for example with business owners, IT operations and information security. Anything still open, we work out with you during the assessment.

  • Application and load profile: user numbers, load peaks, expected growth.
  • Availability target: the downtime your business can tolerate. The availability calculator converts percentages into downtime.
  • RTO and RPO: recovery time and tolerable data loss; see how to set RTO and RPO.
  • Data volume and I/O profile: amount of data and write load of the application.
  • Existing environment: hardware, colocation and virtualization, such as a VMware environment.
  • Timeline: go-live date and possible windows for the switchover.
  • In-house security requirements: for example from your information security management system (ISMS) or from requirements in your industry. The overview of security and compliance responsibilities shows which tasks we take on and which stay with you.

FAQ

Frequently asked questions about design and setup

How long does it take to get a design?

Usually a few business days after your request; you receive an initial assessment within one business day. The preparation list shows what information we need for the design.

How long do build and testing take?

Typically one to three weeks, depending on scope. During this time, we build the cluster, set up monitoring and backups and test failover. You set the migration date.

Can existing hardware be integrated?

We assess that case by case: we check whether and how your existing servers or colocation space can be integrated. They are not required; the cluster can run entirely on infrastructure we manage in German data centers.

Do you test before handover?

Yes. Before handover, we run a full failover test, and you receive the log. During operations, we repeat the tests on a schedule agreed with you.

How much does a managed Proxmox cluster cost?

The published prices apply to the Essentials (€449.99 net per month) and Business (€1,099.99 net per month) packages. We price Enterprise environments by data volume, RPO target and switchover time and state the fixed price in the proposal. The published package prices show what each package includes.

Next step

Your high availability design, in writing and at a fixed price

Tell us about your application, its load profile and your availability target. From that, we derive sizing, redundancy, network and storage, agree on recovery targets with you and set everything out in a design with an architecture drawing, component list and fixed price.

Request a design