Stabilize, then decide

Old Software, No Developer: Rescue a Legacy Application Before You Decide Its Fate

It was written years ago by someone who has since retired, moved on or stopped answering. It still takes the orders. Nobody alive at the company knows how. This week it started throwing errors, and the only manual is a sticky note on the monitor.

November 20, 2026 · 10 min read

orders.acmemachineworks.com · rescue inventory
What could be accounted for on day one
The code turned up. The backups did not.
changes frozen
Code copied off the old serverdone
Logins moved to business accounts4 of 10
Scheduled jobs written down3 of 5
Outside service keys found2 of 4
Database backup that restoresnone found
Illustrative: sample first-day inventory for a fictional machine shop's order system

The short answer

Stabilize before you decide anything. Move every login into business hands, take a full backup of the code and the data and store it somewhere else, and write down how the system runs today. Only then choose between repairing, rewriting or replacing it, with facts instead of fear. A fractional CTO can make that call if nobody inside the company can.

On this page
  1. The machine in the closet that nobody is allowed to reboot
  2. Hour one with software nobody understands
  3. The inventory that ends the guessing
  4. Repair, rewrite or replace, decided once the alarms stop
  5. Six moves that sink a rescue
  6. Who makes the call when nobody on staff can

The machine in the closet that nobody is allowed to reboot

Every company of a certain age has one. A server in a closet, or a hosting account paid on a card that expires next spring, running an application that does one job the business cannot live without: quotes, scheduling, inventory, the customer portal. The person who built it left behind a folder of code, a password in a spreadsheet and a standing instruction to never, ever restart it.

Then it breaks. An error page where the order screen used to be. A report that stopped running. A browser update that made a button vanish. The temptation is to make the big decision on the spot, usually to replace the whole thing with something modern, and to make it while the business is losing money by the hour.

The worst time to choose a new system is during an outage of the old one.

A legacy rescue runs in a different order. First keep it alive. Then make sure that if it dies tonight, nothing is lost. Then understand it. Only then decide what it becomes. Each step makes the next one cheaper, and the decision at the end is made with evidence rather than adrenaline. If the developer is reachable but will not hand over access, start with our page on a developer who disappeared.

Hour one with software nobody understands

  1. 1

    Touch only what is broken

    If part of the system still works, leave it alone. No updates, no cleanups, no reboots for the sake of it. Old systems often run on settings nobody remembers making, and a restart can reveal that a service never comes back on its own.

  2. 2

    Collect the keys

    List every login the system depends on: hosting or cloud account, server, database, domain registrar, code repository, email that receives password resets, and any outside service it talks to. Write down who holds each one today.

  3. 3

    Copy everything off the box

    Back up the code, the database, the configuration files and the scheduled tasks, and store the copy somewhere other than the same server. Then open the copy and check it is complete. A backup on the machine that is failing is not a backup.

  4. 4

    Write down how it runs today

    Operating system, language version, database version, what runs on a schedule, which staff use which screens. Photos of settings pages count. This document is worth more than any line of new code in the first week.

  5. 5

    Find the money paths

    Name the three things that cost money when they fail: taking orders, sending invoices, booking jobs. Those get tested after every change, and they set the priority for everything else.

  6. 6

    Put a watch on it

    Set up an uptime check that alerts a real person, and an alert before the security certificate expires, so the next failure is caught in minutes rather than by a customer.

  7. 7

    Freeze changes

    Until the backup is verified and the map exists, nobody changes anything that is not part of the emergency fix, however tempting the tidy-up.

The inventory that ends the guessing

Once the system is stable and copied, the next job is an inventory. It sounds like paperwork. It is the difference between a rescue and a guess, and it is what any developer, ours or anyone else's, needs before quoting real work.

What to account for in a legacy application, and why each item matters during a crisis
Where it usually hidesWhy it matters now
Domain registrationA registrar account in the old developer's name or emailWhoever is listed as registrant controls transfers; get legal advice if it is not the business
Hosting or cloud accountA personal card, a personal login, an account nobody has opened in yearsLose this and you lose the server, the backups and the logs together
Database and its backupsOn the same server, with a retention setting nobody checkedA restore reaches back only as far as the window set before the problem
Source codeThe server itself, a developer's laptop, a repository under a personal accountNo code, no repair; the running copy may be the only one
Scheduled jobsCron entries, Windows tasks, a script that runs at 2 a.m.Invoices, reports and syncs that stop silently when the server moves
Outside service keysConfiguration files and hard-coded stringsPayment, email and shipping connections fail the moment a key is rotated
Collage of an old beige desktop computer wrapped in cobwebs, with two hourglasses and a daisy, on a pink background
A rescue starts with a list of what exists, not a plan for what comes next.

The domain row deserves a word of caution. Under ICANN's rules, the person listed as the registrant controls transfers, so a domain registered in a former developer's name is the hardest case on this list; that is our reasoning from ICANN's transfer policy, and it is exactly where to talk to an attorney, as ICANN itself suggests. Our page on a domain or DNS emergency covers expiration and DNS. The backup row ties to our database emergency page: Amazon's RDS documentation, as of November 2026, allows a backup retention period between 0 and 35 days, and 0 turns automated backups off. Check what yours is set to today; on an old system, nobody may have looked since the day it was built.

Repair, rewrite or replace, decided once the alarms stop

With the system stable, copied and mapped, the real question can be answered honestly. There are four options, and the right one depends on what the software does, how often it changes and how much of the business runs through it. None of them is always right, and anyone who answers before reading the code is guessing.

Keep it and patch it

When the software does one job well, rarely changes and the fix is contained. Move it to supported hosting, fix the break, document it, and leave it alone. Not glamorous. Often the cheapest honest answer.

Repair and modernize in place

When the core logic is sound but the foundation is aging: an old language version, an unsupported framework. Upgrade in steps on a copy, test the money paths at each step, and keep the business running throughout.

Rewrite it

When the software is central to the business, changes often and fights every change. A rewrite is a project, not an emergency fix. It belongs on a plan with the old system kept alive until the new one has proved itself.

Replace it with a product

When an off-the-shelf tool now does the same job. Plan the data move carefully, because years of records rarely fit a new tool's shape without work, and keep the old system readable for reference.

Rewrites tempt people most in a crisis, because they feel like a fresh start. They also take longest to pay off, and the old system has to keep running the whole time anyway. That is why stabilizing comes first in every case: it buys the time to choose well. When the choice is to build, our custom software development and web app development work picks up from the inventory instead of starting from nothing.

Six moves that sink a rescue

  • Upgrading the language version on the live server to fix one error, and breaking ten other things written for the old one. Our page on a site that broke after a PHP update shows how that goes.
  • Committing to a full rewrite in the first week, before anyone has read the code or listed what it does.
  • Leaving one person as the only holder of the passwords, again, after the last person in that role disappeared.
  • Deleting files and database tables that look unused, which turn out to drive the year-end report.
  • Keeping the only backup on the same server or the same account as the system it protects.
  • Hiring the cheapest bidder to modernize everything at once, with no inventory, no written estimate and no test plan.

Old PHP versions are their own risk, separate from the emergency itself. Our end-of-life PHP page explains what running unsupported versions means for security. If the old system was built on WordPress, the WordPress repair section covers fixed-price fixes for the common breaks, and the failed software update page covers the case where a well-meant update caused the outage in the first place.

Who makes the call when nobody on staff can

The hardest part of a legacy rescue is often not technical. It is that nobody in the company is qualified to judge the options, so the decision defaults to whichever vendor sounds most confident. A fractional CTO solves that: a senior technical person, part time, who owns the decision on the business's side of the table. Our page for non-technical founders explains how that works when the owner is not a developer.

You can steady it yourself if

  • The system is still running and you have every login
  • You can take and verify a full backup today
  • The break is small and someone on staff understands that part
  • There is time to plan before the next failure

Call us if

  • It is down now and the business is losing money
  • Nobody knows where the code or the backups are
  • The only person with the passwords is gone
  • You need someone to read it and tell you the truth about its future

We respond the same business day; hours are on our contact page. Getting a legacy system back on its feet falls under our emergency software repair service, from $575, depending on complexity, with a written estimate before work starts. The emergency pricing page explains what moves that number. Once it is stable, our hosting and maintenance plans start at $55 a month. DataCram is a San Antonio web studio that has been building for the web since 1999 (about). For any other kind of outage, start at the emergency repair hub.

Questions owners ask in an emergency

Stabilize it before deciding anything. Gather every login into business hands, take a full backup of the code, the data and the configuration and store it away from the server, and write down how the system runs today. Leave working parts alone until that is done. Those steps make every later option, repair, rewrite or replace, cheaper and safer.

Sometimes, but rarely as the first move. A rewrite makes sense when the software is central to the business, changes often and resists every change. It is a planned project, and the old system has to keep running until the new one proves itself. Stabilize and inventory the old system first, then compare a rewrite against patching, modernizing in place or replacing it with a product.

We read the code and the server before we quote anything. If it is something we can repair responsibly, the written estimate says so and says what we will do. If it is not, we tell you that plainly and help plan a replacement, including how to keep the old system readable while the new one is built.

The business, not one person. As general practice drawn from AWS's and Google's account guidance, keep account ownership, recovery email and recovery phone on addresses the business controls, and make sure more than one trusted person can reach them. AWS also advises that no single person should hold both the recovery inbox and the phone number for its root accounts.

Yes, if it is done on the live server. Code written for an old version can break in many places at once on a newer one. Upgrade on a copy first, test the screens that take orders and money, fix what breaks, then switch. Running an unsupported version has its own security risks, which our end-of-life PHP page describes.

Emergency work to stabilize and repair a legacy system starts from $575, depending on complexity, with a written estimate before work starts. The estimate depends on how much of the system is documented, whether the code and backups can be found, and how much needs to change. Larger decisions, such as a rewrite, are quoted separately after the inventory.

Keep reading

Free · No obligation

Get the old system steady, then get the truth about it

Tell us what the software does, where it runs and what broke. We reply the same business day with what we would secure first and a written estimate.

Which of these sounds like you? (pick any)

Rather talk now? Call (210) 346-0848.

CallGet a quote