Linux 7.0 PostgreSQL Performance Regression and Developer Impact

PostgreSQL throughput dropping 50% after a kernel upgrade. That’s not a configuration error. That’s a regression that should terrify any team running Postgres in production.

An AWS engineer discovered this while testing Linux 7.0 — the same kernel shipping with Ubuntu 26.04 LTS — and the numbers are genuinely alarming. This affects every team planning to upgrade Ubuntu this year, not just one company’s benchmark.

The short version: Linux 7.0 introduces a scheduler change that cuts PostgreSQL throughput by roughly 50% under certain workloads, and a fix isn’t straightforward. Developers planning Ubuntu 26.04 LTS upgrades need to evaluate this regression before it hits production.

Three things to know upfront:

1. The regression appears tied to Linux 7.0’s scheduler changes, not PostgreSQL itself.
2. AWS engineers first surfaced the issue during internal benchmarking in early 2026.
3. Ubuntu 26.04 LTS ships with Linux 7.0 by default, making this a near-term production risk.

Background: How We Got Here

Ubuntu 26.04 LTS ships with Linux kernel 7.0 as its default kernel, according to the Ubuntu release roadmap documented on Computing for Geeks. That’s a significant jump — and kernel 7.0 brought meaningful changes to process scheduling and memory management, aimed at improving throughput across mixed workloads.

PostgreSQL remains one of the most widely deployed relational databases in the world. The project’s own site describes it as the world’s most advanced open-source database, and that reputation is backed by decades of production hardening across financial systems, SaaS platforms, and cloud infrastructure.

The collision between these two was first documented publicly by a Phoronix report in April 2026, citing an AWS engineer who discovered the performance drop during internal testing. The engineer observed PostgreSQL throughput dropping by approximately 50% on Linux 7.0 compared to Linux 6.x baselines — same hardware, same Postgres version, same workload.

That’s the part that matters. Nothing changed on the application side. The regression is entirely attributable to kernel-level behavior.

Linux kernel upgrades routinely introduce performance shifts. Most are marginal — a few percent here or there. A 50% throughput drop is a different category of problem entirely. The Phoronix report specifically noted that a fix “may not be easy,” which suggests this isn’t a simple patch waiting in a queue.

The timing makes it worse. Ubuntu 26.04 LTS released in April 2026, which means production teams running standard Ubuntu upgrade paths are heading directly toward this issue right now.

What’s Actually Causing the Regression?

The regression appears rooted in scheduler changes introduced in Linux 7.0. PostgreSQL’s process model relies heavily on spawning worker processes and managing inter-process communication — behavior that’s deeply sensitive to how the kernel schedules tasks and manages CPU affinity.

Linux 7.0 modified its EEVDF (Earliest Eligible Virtual Deadline First) scheduler — the successor to the Completely Fair Scheduler — with refinements aimed at reducing latency for interactive workloads. Those changes appear to penalize high-concurrency, process-heavy workloads like Postgres.

Postgres doesn’t use threads. It forks a new process per connection. That architecture, while battle-tested, means it’s more exposed to scheduler behavior than thread-based databases. When the scheduler’s context-switch patterns change, Postgres feels it immediately.

This isn’t speculation. The AWS engineer’s findings directly correlate the performance drop with the kernel version. Rolling back to a 6.x kernel on the same hardware restored previous throughput levels.

Why the Developer Impact Is Broader Than One Benchmark

Teams dismissing this because “it’s just one AWS engineer’s test” are making a mistake.

PostgreSQL powers a significant slice of production infrastructure. According to the Stack Overflow Developer Survey 2025, PostgreSQL was the most-used database among professional developers for the third consecutive year. A 50% throughput regression on the kernel version shipping with the next Ubuntu LTS isn’t a niche problem.

The impact compounds for teams that:

• Run connection poolers like PgBouncer in session mode (more processes, more scheduler exposure)
• Use parallel query execution (worker process spawning scales the problem)
• Deploy on cloud instances where kernel versions are tied to OS images

And the fix timeline is unclear. Kernel maintainers would need to either revert the scheduler change, add a tunable knob, or find a middle path — none of which happens overnight.

Comparing Mitigation Options

The cleanest short-term path is staying on Ubuntu 24.04 LTS with Linux 6.x. Ubuntu 24.04 LTS receives security support through April 2029 — three years of runway while the regression gets resolved upstream. Teams don’t need to rush onto 26.04 LTS on day one.

Pinning a 6.x kernel on a 26.04 base is technically possible but sits outside standard support paths. It’s a workable stopgap for teams that need other 26.04 features, but it requires internal kernel management discipline.

Scheduler parameter tuning is worth exploring for teams with kernel engineering capacity. The EEVDF scheduler exposes some tunables via /proc/sys/kernel/sched_*, and community experimentation will likely accelerate as the regression gets more attention. But this isn’t a realistic path for most application teams.

Three Scenarios Worth Thinking Through

Scenario 1: You’re planning a routine Ubuntu LTS upgrade this quarter.

Stop. Benchmark PostgreSQL throughput on Linux 7.0 before committing to production. Spin up a staging environment on Ubuntu 26.04 LTS, run pgbench against your actual workload, and compare against your Linux 6.x baseline. A throughput drop exceeding 10–15% is a blocker — hold on Ubuntu 24.04 LTS until an upstream fix lands.

Scenario 2: You’re running on AWS, GCP, or Azure with managed kernel images.

Check which kernel version your cloud provider’s Ubuntu 26.04 images ship with. AWS has already encountered this internally, so watch for provider-specific guidance. Some cloud providers may ship modified kernels or issue advisories before broadly rolling out 26.04-based AMIs.

Scenario 3: You’re a platform or DevOps team managing kernel versions for multiple product teams.

This is a coordination risk. If individual teams upgrade OS versions independently without benchmarking, you’ll catch the regression in production under real traffic — not in a controlled test. A centralized kernel policy with mandatory PostgreSQL benchmarking gates before LTS upgrades is the right call here.

What to watch: Phoronix and the linux-kernel mailing list will be the earliest signal for patch progress. The PostgreSQL community may also publish formal guidance. Track both.

What Happens Next

The Linux 7.0 PostgreSQL performance regression is real, documented, and consequential. The data shows:

• A 50% throughput drop on Linux 7.0 versus 6.x, confirmed by AWS internal testing
• The cause appears tied to EEVDF scheduler changes in the new kernel
• Ubuntu 26.04 LTS ships Linux 7.0 by default, making this a live 2026 production risk
• No fix is confirmed or scheduled — teams can’t plan around a patch timeline that doesn’t exist yet

Over the next 6–12 months, expect active discussion on the linux-kernel list, possible scheduler tunables or workarounds from the community, and cloud providers issuing guidance for Postgres-heavy workloads. Ubuntu 24.04 LTS as a holding pattern gives most teams enough runway to wait this out without security exposure.

The regression ultimately comes down to one question: has your team benchmarked Postgres on the kernel you’re planning to ship?

If the answer is no, that’s the only action item that matters right now.

Key Takeaways

• Linux 7.0’s EEVDF scheduler changes cause approximately 50% PostgreSQL throughput loss under high-concurrency workloads
• Ubuntu 26.04 LTS ships this kernel by default — the risk is immediate, not theoretical
• Rolling back to Linux 6.x on the same hardware fully restores performance, confirming the regression is kernel-side
• Ubuntu 24.04 LTS (supported through April 2029) is the safest holding pattern while an upstream fix develops
• Benchmark PostgreSQL on Linux 7.0 in staging before any production upgrade — this is non-negotiable

References: Phoronix — Linux 7.0 AWS PostgreSQL Performance Report | PostgreSQL.org | Computing for Geeks — Ubuntu 26.04 LTS Features

References

1. AWS Engineer Reports PostgreSQL Performance Halved By Linux 7.0, But A Fix May Not Be Easy - Phoroni
2. PostgreSQL: The world’s most advanced open source database
3. Ubuntu 26.04 LTS Features: Kernel 7.0, GNOME 50, What’s New [2026]

Photo by Lewis Kang’ethe Ngugi on Unsplash