Inventory of roles
What each machine carries: directory, files, the database behind Sage, Cegid or Divalto, a web server, a mail relay. Each entry also gets the end-of-support date of its operating system.
Taking over servers always begins with a simple question that is seldom written down anywhere: what runs on these machines, and who relies on it? In a company of forty people the answer rarely fits in three lines. There is a directory, a file share, a database behind the accounting package, a box that pushes orders to the carrier, and a print server whose password nobody remembers. Our first job is to write that list; our second is to make sure nothing on it dies without warning. Your machines may sit in the office, in a rack at a hosting provider or at OVHcloud, and the method stays the same.
We do not start from tooling. We start from the list of services your business cannot trade without, and the alert thresholds follow from it.
What each machine carries: directory, files, the database behind Sage, Cegid or Divalto, a web server, a mail relay. Each entry also gets the end-of-support date of its operating system.
SMART values, RAID array state, power supplies and temperatures read through iDRAC, iLO or the hosting provider console. A weakening disk triggers a warranty claim rather than a wait.
Windows inside an agreed window; Linux with particular attention to the PHP and database versions your website depends on. Always with a way back.
An expired certificate halts a mail flow or a shop as thoroughly as a failed disk. Certificate, domain and licence expiry dates live in a calendar we watch.
Daily review of job results and, at the agreed rhythm, an actual rebuild of one chosen server to prove the copies are good for something.
Who did what, when and why. That single document is what makes NIS2 preparation or an audit run along ISO 27001 lines bearable, and a change of supplier painless.
The first weeks go into understanding what exists. After that the rhythm is familiar to everyone, your own staff included.
Named administrative accounts and an encrypted channel, usually a VPN with two-factor authentication.
Disks, overdue patches, backups, logs. Anything urgent is fixed straight away, the rest goes into a plan.
Probes go in and we decide, service by service, who receives which alert and at what hour.
Fixed windows, a monthly report and a short video call with whoever makes the decisions on your side.
Monitoring that fires four hundred alerts a day has stopped monitoring anything. After a fortnight nobody reads the messages and the real outage drowns in the noise. So we spend time on thresholds, group related alerts together and silence the ones that ask for no action. An alert should mean a human has something to do, right now.
No, and it is usually easier: power, cooling and connectivity are already someone's profession. At OVHcloud, Scaleway or elsewhere we work through the remote console, and when a part fails the physical work goes through the provider's ticket. If the machines stay in your office, what we really ask for is a decent UPS and a management interface we can reach.
Every month it stays in production adds risk, and the argument no longer convinces an insurer or a NIS2 file. We plan the move of its roles onto a supported version or a virtual machine, beginning with whatever migrates painlessly. Until then the box is isolated on the network, its remote access closed and its monitoring tightened.
Patches flagged critical skip the queue: we propose a slot the same day. The rest wait for the monthly window. When CERT-FR warns about a flaw already being exploited in a product you run, we write to you before a slot even exists, with the workaround to apply immediately.
Yes. Passwords live in a vault you hold a key to, administrative accounts are named, and the documentation belongs to you. If you ever change supplier, everything is handed over without negotiation, which is also plain common sense in a GDPR processing agreement.
Tell us what you have and where it runs. We come back with our questions and a realistic monthly figure for running it.
Message received
An answer follows inside one working day. Report an outage that is stopping people working and it moves ahead of everything else.
Nothing here under that name. Check the spelling, or simply choose the nearest large city instead. Since every engagement runs remotely, whichever you pick changes nothing about what we do for you.