Paying R$400 a month for a 3,000-message automation quota while processing over 10,000 messages is the kind of math that pushes a small team toward self-hosting. The alternative that worked here was a Kubernetes cluster where compute time is the only real cost.
Why a small business moved from managed cloud to self-hosting on Kubernetes
Self-hosting on Kubernetes cut more than R$5,000 per month from Erick Wendel's infrastructure costs, according to his June 2025 account of running EW Academy, his JavaScript-focused programming school. The migration replaced per-tool managed subscriptions with an Azure Kubernetes Service (AKS) cluster where he pays only for virtual machine time.
The trigger was a message-quota problem. His marketing team adopted an automation tool priced around R$400 per month for roughly 3,000 messages, but the business was already processing more than 10,000 messages monthly. Scaling that plan was unaffordable, so he looked for open source alternatives he could run himself.
Wendel had previously relied entirely on managed platforms and hosted products, with no self-managed database and no automations. A video player that failed during peak access periods added to his frustration with outsourcing every layer. The decision to self-host came only after the business matured enough that his marketing operation was producing its own tooling demands.
What runs inside the AKS cluster
The cluster runs on Azure's managed Kubernetes, where the control plane costs nothing and the bill reflects virtual machine usage. He operates two virtual machines with 16 GB of RAM each and treats adding a new tool as a zero-cost decision unless it demands heavy storage or data processing.
The workloads are mostly open source tools and the databases they depend on. Node-RED is a low-code automation tool maintained under the OpenJS Foundation, used as the integration hub. Mautic is an open source marketing automation platform that recorded nearly 1 million emails sent during a recent EW Academy launch.
Because several of the open source tools ship with their own database dependencies, the cluster hosts PostgreSQL, MySQL, MongoDB, and Redis side by side. Postgres holds the central data model that replaced the old one-event-in, one-event-out pipeline.
How Node-RED cut R$400 per month from automation costs
The first self-hosted win was Node-RED running in a Docker container. The speaker reports it consumes less than 100 MB per month of resources and eliminated the R$400-per-month automation bill on its own.
Node-RED started as a simple event sink: it received events from external platforms and wrote them to a spreadsheet for the marketing team. Over time it grew into dozens of integrations, including student-progress tracking that sends follow-up emails based on where a learner stopped watching a video.
The savings pattern is the core argument for the migration. Managed automation platforms meter messages and integrations. A self-hosted low-code tool on already-paid VMs meters nothing, so integration count stops being a budget line.
Database modeling and automated backups
Migrating forced a real data-modeling effort, which the speaker describes as the hardest part of the move. Instead of forwarding events between platforms, every student activity is now recorded in his own Postgres database, which required revisiting fundamentals: schema design, business rules, and long-term organization.
Backups run as cron jobs inside the cluster, shipping database dumps to an external location. The stated goal is simple: if the cluster fails completely, the data survives elsewhere.
This is where self-hosting shifts work onto you. On a managed platform, someone else owns the backup routine and the schema constraints. Self-hosted, those responsibilities are explicit tasks, and the speaker says he periodically hires freelancers to set up infrastructure for products he knows less well, update versions, and configure additional alerts.
The disk-full incident: what monitoring caught and what it missed
On June 4, 2025, the production setup failed. A disk alert appeared the night before, was deprioritized, and by early afternoon the disk hit 100%. MySQL went down, and with it Mautic and its dependencies.
Monitoring still did its job. The alert system notified him at least three times before the outage, and recovery took minutes once he acted: a few commands brought everything back online, and he enlarged the disks on other volumes to prevent recurrence.
His conclusion is that monitoring and alerts are essential, not optional, for self-hosting. Critical alerts arrive directly in his Discord channel, and he estimates spending roughly 30 minutes per week on hands-on maintenance, treating that time as a calculated risk rather than a hidden cost.
Is self-hosting worth it without a dedicated infrastructure team?
For small and medium businesses without a large team, the speaker argues self-hosting is worth considering precisely because Kubernetes reduces the micro-management burden. When a container crashes, the platform restarts it and notifies him afterward; he rarely intervenes at that level.
He also pushes back on the idea that managed services save time outright. The micro-problems disappear on managed platforms, but the macro-work, such as configuring reports, backup routines, and business integrations, still requires hands-on effort either way.
The trade-off table looks like this:
Frequently asked questions
- How much did self-hosting on Kubernetes actually save? The speaker reports saving over R$5,000 per month in total, including R$400 per month from replacing one paid automation tool with Node-RED. These are first-hand figures from his June 2025 video, not an independent audit.
- Does Kubernetes itself cost money on Azure? In his AKS setup, the Kubernetes control plane carries no charge; the bill comes from the virtual machines running the workloads. He runs two VMs with 16 GB of RAM each.
- What open source tools are used for automation? Node-RED handles event integrations and low-code automations, and Mautic handles marketing and email automation. Both are open source, and Mautic recorded nearly 1 million emails during one launch campaign.
- What was the biggest operational failure? In June 2025 a disk filled to 100% overnight after an alert was ignored, taking down MySQL, Mautic, and its dependencies. Recovery took minutes, and disk sizes were increased afterward to prevent a repeat.
- What should beginners self-host first? The speaker's own sequence was a low-code automation tool in a container, then databases with automated backups, then monitoring with alerts to Discord. Starting cheap with an isolated lab environment keeps the risk contained while you learn.
Turn your own infrastructure stories into written guides
The lessons in this article came from one developer narrating a year of running his own cluster: the savings, the disk-full outage, the backup routines. If you have knowledge like that sitting in YouTube videos, whether walkthroughs, interviews, or hard-won production lessons, Skala Blog turns that footage into a structured written article. Paste your YouTube URL, get a transcription, and generate a publishable draft you can review before it goes live.
Fork this article
Start a new branch from the same video, shaped your way. You keep the credit; the original keeps the attribution.
A fork in another language is filed as a translation of this article, so the two pages point at each other. You can unlink it later from the editor.
0/240
You are creating
- Format
- For
- Language
- Source
- Your angle
You will be asked to sign in before it is generated.
Buy credits