Client deployment

A safe move of a company system to a new server

The old server no longer received regular security updates, did not encrypt the connection and had no verified data backup. The switchover took 14 minutes, and the running cost fell by about 80 percent.

Client
Jerzy Czupik · continued cooperation after the system recovery
Duration
14 minutes of downtime + a week of supervision

Morning report · 07:00 systems are up

  • Backup 03:12 · restore verified
  • Server up · 42 days without a break
  • Site and system responding · 0.4 s
  • SSL certificate valid · renews itself
  • Updates 3 installed overnight
  • Disk 78% · clean-up scheduled

sample data

Before

  • A server past its standard support period and without regular security updates
  • The connection to the company panel was not encrypted
  • No data backup whose restore had ever been tested
  • A server cost out of proportion to the actual load

After

  • The switchover took 14 minutes and happened outside working hours
  • A daily data backup in another location, verified by a restore
  • An encrypted connection and automatic security updates
  • Running cost lower by about 80 percent

Client testimonial

“The program works perfectly, and on top of that we are saving money thanks to the server change.”
Jerzy Czupik · owner of a company with three points of sale Translated from Polish

The system was sound, but what it ran on no longer was

This is the continuation of the cooperation after the recovery of an abandoned order system. The application was fit for development again, but it still ran on an old, expensive server. The users saw no problem. The owner, however, carried a risk that would only show up in an outage or a loss of data.

  • the operating system no longer received standard support,
  • the connection to the panel was not encrypted,
  • there was no confirmed way to restore the data from a backup,
  • the server’s specification and price far exceeded the company’s needs.

The plan: a move the staff will not notice

I treated the migration like moving a working office. First I prepared the new place, then I copied and checked the data, and only at the end did I change the address under which the system was available. The previous server stayed ready as a way back until the new one had proved stable.

I did not combine this operation with an unnecessary rebuild of the application. The fewer things change during a move, the easier it is to find the cause of any problem.

Switchover in 14 minutes

A few days earlier I prepared the domain for a fast change of address. On a Saturday evening I stopped writes, made the final copy of the database, moved it and checked the most important functions. The whole window lasted from 21:00 to 21:14.

On Monday the team went to the same address, used the same passwords and saw the same system. The old server was not switched off for good until the new one had worked a full week without problems.

A backup that can really be restored

The daily backup goes to storage in a different country than the main server. That way a failure of one site does not take the original and the safeguard at the same time.

A “backup completed” message alone is not enough. I restored the data to a clean database and compared the number of records. Only such a test confirms that, if needed, the company will really get its order history back.

The week of supervision revealed one more thing

After a few days the staff reported that the list of orders opened slowly. The system fetched the whole multi-year history every time. Limiting the default range and compressing the transferred data cut the opening time from several seconds to about half a second. The fix fell within the agreed week of supervision.

The client remains the owner

After the supervision ended, the server account, the payments and full control were handed over to the client’s company. I kept only the technical access needed for further care. Changing a contractor should not mean losing access to your own system.

The result

  • 14 minutes of planned downtime outside working hours.
  • An encrypted connection and an automatically renewed certificate.
  • A daily backup off the main server, confirmed by a test restore.
  • The list of orders opening in about half a second instead of several seconds.
  • Server cost lower by about 80 percent while keeping the performance needed.
  • The account and the billing for the server on the client company’s side.

The times and the saving given here concern this particular project. They come from measuring the maintenance window and the list’s opening time, and from comparing the bills for both servers.


Because of a confidentiality agreement I leave out the company’s name, its location, staff data and the addresses and parameters of the infrastructure.

Technologies in this project

Hetzner CloudUbuntu 24.04CaddyLet's EncryptDockerrcloneMySQLPM2

A similar situation at your company?

Call and tell me what has to work.

You do not need to know the technical terms. The first conversation is free — I will tell you where I would start.

Before you pay, you see how it will work.

A free demo: a working demo, a mock-up, a cost calculation or a repair plan. Then a fixed price in writing and a warranty of at least 30 days on every stage.

If I don’t pick up, I call back the same business day.

← All case studies