Ryz Labs/Migrations/Python 3.10 end of life
Migrations

Python 3.10 End of Life: Upgrade Plan for Production Services

Python 3.10 is frozen. Here is how to find every interpreter still running it, what breaks between 3.10 and 3.14, and a staged plan to cut over safely.

Python 3.10 reached end of life on October 1, 2026, according to the official Python version status page. It no longer receives security fixes from the project. Our recommendation is to move production services to Python 3.14, which is still in its bugfix phase and is supported until October 2030, testing first on 3.12 to flush out the distutils and imp removals that cause most of the breakage.

Key dates

Dates come from the official CPython release calendar and each version's PEP, checked November 21, 2026.

VersionReleasedEnd of lifeStatus on November 21, 2026
Python 3.10October 4, 2021October 1, 2026End of life (3.10.22 was the final release)
Python 3.11October 24, 2022October 2027Security fixes only
Python 3.12October 2, 2023October 2028Security fixes only
Python 3.13October 7, 2024October 2029Security fixes only since October 1, 2026
Python 3.14 (recommended)October 7, 2025October 2030Bugfix releases with binaries, planned until October 5, 2027

Notice the trap in that table. Jumping from 3.10 to 3.11 or 3.12 buys you a single year or two before you repeat this exercise. Python 3.13 just left its bugfix phase on October 1, 2026, the same day 3.10 died. Python 3.14 is the oldest line that still ships regular bugfix releases with installers, and it carries the longest runway of any release a cautious team would put in production today. Python 3.15 shipped on October 9, 2026 and runs to October 2031, but most teams will want their dependencies to publish wheels for it first.

What end of life means for your systems

PEP 619, the 3.10 release schedule, says it plainly: as of October 1, 2026, the 3.10 codebase is frozen, no further updates will be provided, and no issues of any kind will be accepted on the bug tracker. That goes beyond what most teams expect. Python 3.10 had already stopped getting binary installers in April 2023; since then it lived on source-only security patches. Now even those have stopped. If a vulnerability is reported in the 3.10 standard library tomorrow, the CPython team will fix it in supported branches and leave 3.10 alone.

Three practical effects follow.

We are deliberately not claiming that your libraries will stop working on 3.10 on a specific day. Each package sets its own policy. What we can say from the Python docs is that the standard library you are running will not change again.

Find your exposure

The interpreter version hides in more places than the production Dockerfile. Sweep all of these before you estimate anything.

To speed this up, paste a Dockerfile, pyproject.toml or requirements file into our free EOL deadline radar. It returns a dated report of every runtime it recognizes and how long each one has left.

What changes in the upgrade

Going from 3.10 to 3.14 crosses four "What's New" documents. These are the changes we see bite production code most often, each taken from the official Python documentation.

distutils is gone (3.12)

Python 3.12 removed the distutils package, which PEP 632 deprecated in 3.10. Any setup.py that imports from distutils.core, and any runtime use of distutils.version.LooseVersion or distutils.util.strtobool, fails with ModuleNotFoundError. The docs note that installing setuptools still provides distutils for projects that cannot be updated, which makes a useful bridge but not a destination.

venv no longer installs setuptools (3.12)

Also in 3.12, ensurepip stopped bundling setuptools and venv stopped installing it. Code that calls pkg_resources at runtime, a common pattern for reading package versions or entry points, will fail in a fresh virtual environment unless setuptools is declared as a real dependency. The cleaner fix is importlib.metadata.

The imp module and old asyncio coroutines (3.11 and 3.12)

Python 3.11 removed the @asyncio.coroutine decorator; generator-based coroutines must become async def. Python 3.12 then removed imp entirely, with a mapping table in the docs: imp.reload becomes importlib.reload, imp.find_module becomes importlib.util.find_spec, imp.load_source needs a short importlib recipe. Plugin loaders written years ago are the usual hiding place.

Network and TLS helpers removed (3.12)

ssl.wrap_socket() and ssl.match_hostname() were removed in 3.12. The release notes are blunt that wrap_socket neither sent SNI nor validated the server hostname. The replacement is an ssl.SSLContext and its wrap_socket method. asyncore, asynchat and smtpd went at the same time; the docs point to asyncio and the aiosmtpd package from PyPI.

Dead batteries removed (3.13)

Python 3.13 completed PEP 594 and removed 19 legacy modules, including cgi, cgitb, crypt, telnetlib, pipes, imghdr, uu and nntplib, along with 2to3 and lib2to3. The docs list replacements: urllib.parse.parse_qsl or the email package for cgi form parsing, hashlib, bcrypt or argon2-cffi for crypt, shlex.quote for pipes.quote, and PyPI redistributions such as standard-cgi when you need time.

Test suite and parsing details (3.11 and 3.12)

Python 3.12 removed the old unittest aliases such as assertEquals, failUnless and assertRaisesRegexp, so a passing suite on 3.10 can error out before it runs a test. Python 3.11 dropped the 'U' file mode, made random.sample() reject sets, and restricted inline regex flags like (?i) to the start of a pattern. In 3.12, an invalid escape such as "\d" in a non-raw string now raises a SyntaxWarning, and the docs say it will become a SyntaxError in a future version.

Process and event loop defaults (3.14)

Python 3.14 made forkserver the default multiprocessing start method on Linux and other non-macOS Unix systems, replacing fork. Workers that relied on inheriting module globals can hit NameErrors or pickling errors. In the same release, asyncio.get_event_loop() raises RuntimeError when no loop is running instead of creating one, and the docs recommend asyncio.run(). Celery-style worker pools, Gunicorn hooks and scripts that start loops by hand deserve a close look.

The payoff

The upgrade is not only cleanup. The 3.11 notes report that 3.11 is between 10% and 60% faster than 3.10, with a 1.25x average on the standard benchmark suite. You also gain tomllib, exception groups, much better error locations, PEP 695 generics syntax, lazily evaluated annotations, and free-threaded builds that are officially supported (but optional) as of 3.14. Measure your own workload rather than assuming those numbers.

The upgrade plan

This sequence works for a single Django or FastAPI service and scales to a monorepo of workers and scripts.

Inventory and baseline

Turn the exposure sweep into a list: every service, job and function with its current interpreter, owner and deploy path. Record a performance and error baseline on 3.10 (latency percentiles, memory, worker throughput) so you can prove the new version is at least as good. Generate a fresh lockfile and note which pinned packages lack wheels for 3.12 and 3.14.

Compatibility testing

Run the existing suite on 3.10 with warnings promoted to errors: python -W error::DeprecationWarning -m unittest, or the equivalent filter in your test runner. Every warning that fires is a future removal. Then add 3.12 and 3.14 to the CI matrix side by side with 3.10. The 3.12 leg catches distutils, imp, ssl and unittest breakage in isolation; the 3.14 leg catches the multiprocessing and asyncio changes. Upgrade or replace dependencies until both legs are green.

Staged rollout

Build a new image on python:3.14 and run it alongside the 3.10 fleet. Start with internal tools and batch jobs, then a canary slice of web traffic behind your load balancer or service mesh. Compare error rates, p95 latency and memory against the baseline. Watch process pools in particular, because forkserver changes how workers start.

Cutover and rollback

Shift traffic in steps. Keep the 3.10 image tagged and deployable until the new one has carried full load through a normal business cycle, including nightly jobs and month-end batches. Because nothing about the data layer changes, rollback is a redeploy of the old image, which is why this upgrade can be done calmly.

Cleanup

Remove 3.10 from CI, bump requires-python, delete compatibility shims and the setuptools-for-distutils bridge, and retire the old base images from your registry so nobody rebuilds them. Add the next deadline to your calendar; our Python 3.11 end-of-life plan covers teams that are one version ahead.

PhaseGoalKey techniqueExit criteria
Inventory and baselineKnow every 3.10 interpreterRepo, image and cloud runtime sweepOwned list plus metrics baseline
Compatibility testingGreen tests on 3.12 and 3.14Warnings as errors, multi-version CI matrixNo failures, no deprecation warnings
Staged rolloutProve behavior under real loadSide-by-side images, canary trafficCanary metrics match baseline
Cutover and rollbackAll traffic on 3.14Stepped traffic shift, old image kept warmFull business cycle with no regressions
CleanupNo path back to 3.10Remove matrix legs, shims and old imagesZero 3.10 references in repos and registries

Risk matrix

RiskLikelihoodImpactMitigationEarly warning
A pinned C-extension package has no wheel for 3.14HighHighUpgrade the pin, or ship on 3.13 first and revisitpip tries to build from source in CI
setup.py or runtime code still imports distutilsHighMediumMove to setuptools or packaging; use importlib.metadataModuleNotFoundError on the 3.12 CI leg
pkg_resources missing in fresh virtual environmentsMediumMediumDeclare setuptools or replace with importlib.metadataImport errors only in newly built images
Worker pools break under forkserverMediumHighMake worker state picklable, or set the start method explicitlyNameError or PicklingError in worker logs
Hand-rolled event loops fail on get_event_loop()MediumMediumRefactor entry points to asyncio.run()RuntimeError at process start
Code depends on a removed dead-battery moduleMediumMediumUse the documented replacement or PyPI redistributionImport failures on the 3.13 or 3.14 leg
A forgotten Lambda or cron host stays on 3.10MediumHighPer-account runtime queries plus host checksEOL radar or scanner still reports 3.10

Typical team shape

A 3.10 exit is mostly careful, unglamorous work spread across many repos, and the people who do it well have seen these removals before. The usual shape is a tech lead who owns the inventory and rollout order, two or three senior Python developers who fix the compatibility breaks and replace dead modules, and a DevOps engineer who rebuilds base images, CI matrices and canary routing. Teams with heavy infrastructure work often pair this with our DevOps consulting group.

Many teams begin with a pod of 3 to 5 senior engineers (a tech lead plus engineers), and a full pod is about seven. Ryz engineers on your team start from $7,000 per month (our published range is $7,000–$15,000 per engineer per month, mostly mid-level to senior), quoted per team. Once you say go, engineers can typically start in under two weeks. Ryz engineers work on your team, reporting to your leads, in your repos and on US business hours. If you would rather keep the work in-house and only want a second pair of eyes on the plan, that is a fine choice too. When you are ready, scope your team with us.

Related

FAQ

Is Python 3.10 still safe to run after end of life?

It still runs, but it no longer gets fixes. PEP 619 says the 3.10 codebase is frozen as of October 1, 2026 and 3.10.22 was the final release, so any new vulnerability in its standard library stays open in your builds. Treat each remaining 3.10 service as a tracked risk with an owner and a date until it is upgraded.

Which version should we upgrade to?

Python 3.14 for most teams. It is still in its bugfix phase with regular binary releases and is supported until October 2030. Python 3.13 is a reasonable fallback if a critical dependency lacks 3.14 wheels, though it moved to security-only fixes on October 1, 2026. We would avoid 3.11 and 3.12 as targets because their remaining support is short.

How long does the upgrade take?

It depends on how many services and images still run 3.10, how much code uses distutils, imp, pkg_resources or removed modules, how many C-extension dependencies need new wheels, how good your test coverage is, and how much multiprocessing and hand-built asyncio code you have. An inventory and a 3.12 CI run usually tell you most of what you need to estimate it.

Can we pay for extended support?

Not from the Python project, which provides no further updates for 3.10. Some operating system and platform vendors maintain their own patched Python builds on their own schedules, so check with your Linux distribution or cloud vendor. Treat any such option as a bridge while you upgrade, not a reason to stay.

What does it cost to bring in engineers?

Ryz engineers start from $7,000 per month, and our published range is $7,000 to $15,000 per engineer per month, mostly mid-level to senior. We quote per team, so the total is team size times duration times the monthly rate, and you see the plan, the price and the names of the people before work starts.

Questions we didn't answer? Email info@ryzlabs.com.

Thanks — your message has been sent. We’ll get back to you soon.

Something went wrong while sending your message. Please try again or email info@ryzlabs.com.

Explore Ryz Labs

Software development servicesAI development companyIndustriesStaff augmentationAI pod teamsNearshore software developmentHire engineers by roleEngineer ratesGuidesBuyer guidesCase studiesHow we vet engineers

Ryz Labs

Senior engineers on your team. AI pod teams that ship.

Tell us what you're building. You get a scoped plan, a price and the names of the people who would do the work.

  • Only the top 1% of tens of thousands interviewed make it
  • On US business hours, including New York hours
  • Trusted by Fortune 500 engineering teams

Tell us who you need

A Ryz partner replies with a scoped team plan.

Start a conversation →

Prefer to talk? Book a 15-min call →