EOL (end of life) means that the vendor has stopped supporting a piece of software or hardware: it no longer releases updates, bug fixes or – most importantly – security patches for it. An EOL operating system will still boot and run, but every newly discovered vulnerability in it stays unpatched.
Technology teams know they have to manage the software lifecycle and avoid running software that is no longer supported. Yet organisations often get into trouble because of legacy systems – sometimes through no fault of their own.
In this article we clarify the terms, list which widely used systems are reaching end of support around now, and then go through the strategies for handling EOL systems. The examples are about operating systems, but the advice applies to any technology reaching the end of its lifecycle, whether hardware or software.
EOL, EOS, ESU: what’s the difference?
For software, the terms EOL and EOS (end of support) are mostly used for the same thing: the day after which there are no more fixes. Hardware vendors, however, have several stages: after the EOL announcement, the product first stops being sold (end of sale), and only later stops being serviced (EOSL, end of service life). There is also an important milestone before the end of support, and even after it expires there may still be a paid add-on:
| Term | Meaning | What do you still get? |
|---|---|---|
| End of active (mainstream) support | No more new features or non-security bug fixes | Security patches, yes |
| EOL / EOS | End of vendor support | Nothing; the vendor no longer fixes the product |
| EOSL | For hardware, the end of servicing and spare parts supply | At most from a third party |
| ESU (Microsoft) | Paid Extended Security Updates | Security patches only, for up to three years |
| ESM, ELS, ELTS | Extended support for Linux: Ubuntu Pro (ESM), TuxCare (ELS), Freexian (ELTS for Debian) | Security patches, often for a narrower set of packages |
You can quickly check the EOL date of any software on endoflife.date, which collects the lifecycles of hundreds of products in one place.
Which systems are reaching end of support around now?
Some widely used systems, as of November 2026:
| System | End of support | Note |
|---|---|---|
| CentOS Linux 7 | 30 June 2024 | No version of CentOS Linux receives fixes any more |
| Ubuntu 20.04 LTS | 31 May 2025 (standard support) | Until 2030 with Ubuntu Pro (ESM) |
| Windows 10 | 14 October 2025 | After that, paid ESU only |
| PHP 8.1 | 31 December 2025 | PHP 7.4 reached end of life back in November 2022 |
| Debian 11 “Bullseye” | 31 August 2026 (LTS) | Paid ELTS for some packages |
| PHP 8.2 | 31 December 2026 | Now receives security fixes only |
| Windows Server 2016 | 12 January 2027 | After that, paid ESU only |
| Ubuntu 22.04 LTS | May 2027 (standard support) | Longer with Ubuntu Pro |
What does end of support mean?
To keep up with technological change, most software gets a major update every few years. Users are expected to move to the newer releases, but they need time to do so – especially when it comes to something like an operating system.
That is why vendors keep supporting a release for a while even after a newer version has replaced it: this leaves time for testing and migration.
At some point, however, the vendor has to stop supporting older releases. It isn’t realistic, for example, to keep supporting a 30-year-old operating system.
Software like this, which the vendor no longer supports or maintains, is said to have reached its end of life (EOL). This means the vendor no longer provides updates, bug fixes or technical support, and – in most cases – there will be no critical security updates either.
So don’t count on a patch from the vendor if a serious security flaw is found in an operating system at the end of its lifecycle. Instead, the vendor expects users like you to have moved to a supported version of the operating system by then.
As a result, anyone relying on outdated software without vendor support is exposed to security holes, since newly discovered vulnerabilities no longer get fixed.
Why is EOL software a problem at all?
The term “end of life” says it all, and sysadmins don’t need any more motivation to upgrade to the latest version of an operating system before it’s too late. In practice, technical teams also have plenty of time to find out, because software vendors publish their support schedules well in advance: Microsoft, for example, set the 14 October 2025 end of support for Windows 10 on its lifecycle page years beforehand.
So where does it go wrong? Why is EOL software used so often despite the risks?
There are several reasons. EOL software is deceptively stable, since everything works. Why throw out a solution that serves its purpose flawlessly just because of a few perceived security concerns?
Upgrading a server to the next-generation operating system is cumbersome, to say the least, so many sysadmins put the process off. Sometimes there simply aren’t enough people for the migration.
Many are also wary of the risk: just because the current version runs smoothly doesn’t mean the latest one will work without problems. The software currently installed may not work with newer versions, and configuration changes can cause outages.
Nevertheless, moving to a supported, patched release of the operating system is a primary task for technical teams, and in most cases it can and should happen in time. We have collected the risks in detail in the article 5 risks of using unsupported software.
CentOS as a cautionary tale
We said “in most cases” because when an organisation gets into trouble with EOL software, it isn’t necessarily the fault of overworked or careless sysadmins. There can be many reasons behind the practical consequences of running outdated software, including the vendor’s own decisions.
This is what happened with CentOS. CentOS is a widely used Linux distribution based on Red Hat Enterprise Linux (RHEL). People liked CentOS because it was a free, open-source alternative to RHEL that had proved itself in data centres and other enterprise environments.
In December 2020, the Red Hat-sponsored CentOS project unexpectedly announced that it was shifting its focus to CentOS Stream, and that support for CentOS Linux 8 would end much earlier than expected, on 31 December 2021. For CentOS 8 users, this opened up a sudden gap.
The announcement left many organisations without a stable and reliable operating system: to get vendor support back, they either had to spend significant sums on RHEL, or test an alternative such as AlmaLinux or Rocky Linux and then migrate to it.
Even after this, many organisations kept running their outdated CentOS 6 or CentOS 8 machines without vendor support. CentOS 7 followed the RHEL 7 lifecycle and also reached its end of life on 30 June 2024, so since then no version of CentOS Linux has received security patches. This matters because organisations that had no good strategy for software reaching its end of life ended up in a very difficult position.
Why it matters: security
Let’s look more closely at why running an EOL operating system is such a big problem. By far the biggest danger of using an EOL operating system comes from vulnerabilities discovered after support has ended. Why? Because the vendor doesn’t release patches.
Linux kernel version 2.6.32, for example, has received no fixes since 2016, yet many vulnerabilities have been found in it since then, even in 2019. Linux servers built on this old kernel may therefore be vulnerable to denial-of-service (DoS) attacks, for example.
If an EOL operating system stays in service, it receives no more security patches, so every vulnerability that becomes public makes your server an open target. Without patches, sysadmins cannot protect the infrastructure.
An attack is also more likely than you might think: since many servers are publicly accessible and attackers use automated scanners, there is a real danger that sooner or later they will find out which servers are vulnerable.
Attackers essentially scan and map your systems 24 hours a day, and pounce on unsupported, vulnerable operating systems. The consequences can be serious.
Why it matters: compliance risk
Regulatory requirements related to financial and health information prescribe specific cybersecurity procedures that protect customer data. Running unsupported, unpatched software runs directly counter to the requirements of many of these compliance standards.
One such requirement, for example, is patching critical vulnerabilities within a set time. But what if you can’t get the patch? Similarly, compliance regulations may stipulate that the organisations concerned cannot use software that isn’t covered by support from the vendor or a third party.
PCI DSS requirements, for example, apply to companies that handle payment card data, and specifically require patches for critical vulnerabilities to be installed within one month of their release. Organisations that don’t meet this deadline are not PCI DSS compliant.
Outdated software can lead to heavy fines and protracted lawsuits that can drag on for years after a data breach: the lawsuits and settlements over the 2013 Target data breach, for example, went on for years.
Why it matters: other reasons
If the cybersecurity and compliance concerns about running end-of-life software aren’t alarming enough, we suggest you also consider the following reasons:
- Incompatible software: once an operating system is no longer supported, developers of third-party applications also drop support for the old system. Newer versions of applications may work incorrectly on the old operating system, or simply not start at all.
- Poor performance: it isn’t unusual for older software to run on older hardware, so the bottleneck is often the outdated infrastructure itself. If you rely on an outdated operating system, you also miss out on the performance improvements of the latest operating systems.
- Reliability: since older software is no longer supported, nobody fixes crashes and bugs either. If the software fails, your server may not even boot, which affects SLAs and availability.
- High costs: constantly patching software the vendor no longer supports quickly becomes expensive, and you never know when an unexpected bug will cost you dearly. Vendors typically charge a per-device premium for extended support – Windows 10 Extended Security Updates (ESU), for example, cost businesses $61 per device in the first year, and the price doubles every year – which can be very expensive for a large organisation.
So yes, EOL operating systems may seem to work fine, but in reality they are much riskier, and there’s a good chance you’ll hit a wall sooner rather than later. As a start, it’s worth assessing where you are exposed to the risk of outdated software.
Start by assessing the current situation
If you don’t have an inventory strategy yet, now is the time to create one. It isn’t only about software: hardware can also reach the point where it’s time to retire. Inventory management helps you determine which infrastructure and software need upgrading, and which infrastructure needs retiring.
The inventory process shows which software has reached its end of life, and lets IT departments quickly establish which software – if any – the vendor no longer supports, and make plans to upgrade or replace it as soon as possible.
Based on the inventory, you can identify which machines or nodes run an operating system close to its end of life – and where you still have plenty of time to deal with the EOL problem. The inventory should also show which of the operating system instances at (or close to) end of life are the most critical, so you can prioritise the work of removing, decommissioning or replacing instances.
By introducing an inventory process, organisations can plan the efficient replacement or upgrade of outdated software, reducing the risk of unexpected downtime or security holes.
Migrate as fast as you can – if you can
Postponed software and operating system upgrades can pile up quickly. The migration should be done as quickly as possible, to avoid a chain reaction of delayed upgrades – and a rushed migration that can end up breaking things.
That is why planning is critical. With a thorough inventory in hand and working with the vendors’ EOL schedules, technical teams should be able to schedule the migration so that it reduces the pressure but the work still gets done in time.
Business-critical infrastructure has to be upgraded, but the new operating system should always be tested first. A test environment that mirrors production (staging) can help catch migration problems you couldn’t have foreseen.
Decommissioning is also an option: if you aren’t going to upgrade a server, it may be time to scrap it. An alternative can be moving the services of decommissioned machines to the cloud or to a virtualised environment.
A migration like this can also be done in a clustered environment without stopping services: our case studies include the step-by-step migration of unsupported CentOS 6 servers to AlmaLinux 8, and on the hosting side, the move from CloudLinux 6 to 8.
Last resorts
So yes, sometimes it just doesn’t work. You can’t migrate fast enough, there’s no extended support available, or for some reason the workload simply requires you to run an outdated operating system as it is, whether you like it or not.
Before you turn to isolation, check whether extended support is available: Microsoft offers ESU, Canonical offers ESM in Ubuntu Pro, TuxCare offers ELS for CentOS and CloudLinux, and Freexian offers ELTS for Debian. These cost money, but they give you time to migrate while security patches keep coming.
Legacy applications: the example of PHP
EOL doesn’t only affect operating systems. An old PHP application often won’t run on a newer PHP, and rewriting it can take months – meanwhile, the old PHP version (for example 7.4 or 5.6) hasn’t received security fixes for years, and if it runs on the same server as other applications, it holds the whole server back.
Our own open-source project, PHP EOL, was built for this situation: PHP-FPM and CLI Docker images for PHP versions 5.2–8.2, with the backported security fixes of CloudLinux alt-php, based on AlmaLinux 9. The old application keeps running in an isolated container, while the rest of the server can use a supported operating system and PHP until the rewrite is finished. The source code and the images are available on GitLab.
Isolation
If you absolutely have to run an unsupported, unpatched and potentially vulnerable operating system, you should think in terms of isolation and risk management:
- Network isolation: use a separate network so that machines running the unsupported operating system can’t communicate with other machines. If network segmentation cuts them off from the rest of the devices on the network and from the internet, this often protects EOL devices from attacks – although this isn’t a perfect solution either, and it comes at the expense of usability.
- Virtualisation for isolation: if the outdated operating system runs in a virtual machine, it is easier to keep under control – in the event of a security incident it can be quickly restored from a clean disk image, and it’s easier to limit what it can reach in the outside world. A compromised machine can be quickly disconnected and restarted.
- Application control and allowlisting: like the previous suggestions, this strategy also isolates the vulnerable operating system by only allowing known “good” applications to run – and to communicate with the system. This model denies all access by default and only permits pre-approved connections.
Nevertheless, running an unsupported operating system in isolation is undoubtedly a risky strategy – there are countless attack vectors, and there is a chance that a cunning attacker will defeat the isolation efforts with a little lateral movement.
Conclusion
With proper planning, you don’t have to resort to desperate measures such as isolating machines that run an unsupported operating system. Ideally, the team migrates in time and always runs a supported operating system for which the vendor regularly releases patches.
Life can get in the way, however – even through hasty decisions by vendors, as the example of CentOS shows.
We also help with assessing and migrating outdated servers and with keeping legacy applications running securely; you can read about this on the Infrastructure and operations page.