What does website maintenance involve?
A website does not wear out, but everything around it changes: library versions, certificates, browsers, Google's rules, providers moving offices. Maintenance means following those changes before they break something: regular updates verified by a real build, monitoring, backups when there is data, bug fixes and small changes.
Antiquité Mallarmé is a good example: a first site in 2016, an application in 2020, a rewrite in 2024, a fourth version in 2026, and continuous additions in between (customer reviews, pages, cookie consent, a form API). The same developer knows the history and the reasons behind each choice.
The infrastructure I run
My projects run on OVH servers, behind a shared Traefik proxy that obtains and renews Let's Encrypt certificates automatically. Each site or API is a Docker container described by a common file and one file per environment (development, staging, production), deployed by GitLab CI after its image is built and published.
Non production environments answer with a header that forbids indexing, secrets are written to the server by the pipeline and never stored in code, and every merge request goes through a three part automatic check: lint, tests, typing.
Locally, the same proxy issues browser trusted certificates for every project with no manual step: development runs in the same conditions as production, HTTPS included.
Application maintenance: taking over existing code
You have an application built by someone who is no longer available? I start with an audit: code, dependencies, tests, deployment. I give you an honest assessment with risks ranked by severity, then we decide together what deserves fixing, upgrading or leaving as is.
I move fastest on Python (Django, FastAPI) and React projects, containerised or to be containerised. I also take over applications written in other languages, Java, .NET or PHP: the audit is there to measure the gap, and I price it before committing.
Fractional CTO: who is it for?
For a company or a charity with technical questions but no technical team: choosing a provider, reviewing a quote, deciding between two solutions, preparing a hire, writing specifications, taking back control of domain names and accounts. I play that role a few days a month, with the eye of a developer who has shipped.
For TapToSupport, a donation app for charities, that work came before the code: a comparison of four payment providers, a reasoned choice of Stripe Connect, and an analysis of the constraints of tap to pay on phones, which put Android first.
What a month of maintenance looks like
Every update goes on its own branch. The pipeline checks linting, tests and a complete build of the site, then the change is deployed to staging before production. A complete build matters more than a type check alone: some defects only appear in a production build, and I have met several that no development server would have revealed.
After each release, I check from the outside what a visitor or a search engine sees: pages answer without useless redirects, old addresses redirect properly, staging carries its noindex header and production does not, certificates are valid. Providers named in the legal notice are checked again in the company registry before those pages change. You receive a short report of what changed and what remains to watch.
The person who wrote the code answers
In classic maintenance, the developer who wrote the code is often gone, and every request starts with a rediscovery phase. With me, the person who answers is the one who made the decisions, or documented them during the takeover audit. The conventions I apply are the same across all my projects: same deployment layout, same verification pipeline, same typography and legal page rules. A fix found on one site benefits the others, and nothing depends on individual memory: everything is written down.
Hosting and managed services: what I was already running in 2016
Before the sites and APIs I deliver today, Business Agile first sold operations. Between 2016 and 2018 some thirty clients were hosted or managed by us, and that period formed the reflexes I still apply: one environment per stage, verified backups, and a server configured for what it carries rather than with generic settings.
The work covered the whole range. For TV ONLY, level 3 managed services across several servers, with backups installed at the client's premises, the move from Apache to nginx and from PHP in CGI mode to PHP-FPM, a certificate rollout and a fix for memory over-use on a database server. For Coding Days, a virtual server and its nginx, PHP and MySQL stack. For Talk Talk World, managed services on the SpeakShake servers and duplication of the platform to its two variants. For EVEZ, two servers behind a load balancer.
Other relationships were simpler and lasted: Check and Visit, in the company's early days in Rennes, Cadeaux Pourris, Dual Cat, Savi Photographie, Plisson Communication, Changement Carte Grise, Mauvais Oeil Graphics, and Dettachée de Presse, whose server was tuned for dense traffic. Hosting, domain names, DNS zones, mail: subjects with no glory, but they are the ones that break on a Sunday.
I no longer resell shared hosting. What I kept from those years is the part that matters: knowing how to configure and run the infrastructure my code sits on, instead of subcontracting it and hoping.
My limits
I do not offer 24/7 on call support with a contractual response time of a few minutes. If your business depends on that, you need an operations team, and I can help you choose one.
