DEV Community: Upsun

Stop waiting for preview environments 🚀

Flora Brandão — Fri, 30 Oct 2026 16:08:19 +0000

Opening a pull request and waiting hours for someone to set up a testing environment kills developer momentum. Code diffs rarely catch full-stack bugs like database migration conflicts or layout glitches.

Here is how zero incremental configuration per branch changes the setup:

Every new pull request automatically spins up a complete preview environment with your app code, database copy, and search index.
One project-level YAML file sets up your architecture once, so every branch inherits its own environment automatically.
Inactive environments automatically pause after 14 days to prevent idle compute waste.

Getting a live URL the moment a PR opens keeps code reviews moving fast without manual staging work.

Check out the full article to see how zero-config previews work in detail:

Full-stack preview environments with zero config | Upsun

Open a branch and get a full working environment with code, database, search index, files and a shareable URL. No tickets, no per-branch config needed

upsun.com

Debugging production only bugs with preview environments 🐛

Flora Brandão — Fri, 30 Oct 2026 10:17:48 +0000

Reproducing a bug that only fails for specific annual accounts in production is a massive grind, especially when all your local tests pass.

Here is how to fix it without risking sensitive data:

Branch from main to clone production services, configuration, and data into an isolated environment
Use sanitization scripts in your deploy hook to scrub personal details while preserving the exact billing state
Open SSH tunnels and attach a remote debugger to step through logic against real failing accounts
Validate the fix in the same environment where you diagnosed the bug before merging

Testing fixes against the actual conditions that caused the failure removes the guesswork from debugging.

Check out the full article to see the complete step by step workflow:

From bug report to verified fix in one workflow | Upsun

Follow one bug from ticket to fix: clone production, reproduce it with real data, debug it, and validate the fix before it reaches your customers

upsun.com

[NEWS] Object storage is now built into Upsun Cloud 🚀

Flora Brandão — Thu, 29 Oct 2026 15:31:11 +0000

If you've been running a separate object storage provider alongside Upsun to handle large asset libraries, you no longer have to.

Object storage is now built into Upsun Cloud. You declare it like any other resource; your app reads and writes through the standard S3 API, and it shows up on your existing invoice. No extra account, no separate integration, no additional bill to reconcile.

It supports asset-heavy applications like Pimcore, Shopware, and Ibexa, as well as media, user uploads, AI artifacts, and exports.

Turn it on from the Console or with one CLI command:

upsun resources:set --object-storage myapp:524288

upsun.com

Scaling slow web applications: Decision charts for resource bottlenecks 📊

Flora Brandão — Thu, 29 Oct 2026 10:04:35 +0000

Having a slow web application without knowing what to scale is a major engineering grind. Guessing whether to add replicas or adjust worker processes usually leads to wasted resources.

Here is how to analyze web application performance:

The problem: Applications slow down, but it is hard to tell if the application or the database is struggling under CPU, memory, and disk pressure.
The fix: Use decision charts to trace where requests spend time, pinpoint bottlenecks, and decide on adjustments like scaling replicas or worker counts.
Why it matters: You can make data-backed scaling moves based on real metrics and verify whether your changes actually improved speed.

Check out the full technical write-up to see the complete decision charts and diagnostic steps:

So, what to scale next? - Upsun Developer

Your web application is slow. Use these decision charts to connect CPU, memory, and disk pressure to your next scaling move, from replicas to fewer workers.

developer.upsun.com

Why faster code generation Is flooding your review queue ⚡

Flora Brandão — Wed, 28 Oct 2026 11:19:47 +0000

Most engineering teams adopted AI coding assistants expecting overall delivery velocity to skyrocket. Instead, developers are writing code faster while team-level delivery stays completely flat.

The real grind is that writing code was historically the primary bottleneck in the software development lifecycle. Because of that constraint, engineering teams designed review and approval processes around human-paced authorship. When you introduce AI coding tools, code gets generated rapidly and becomes cheap to change, but the downstream review pipeline remains unchanged. The constraint did not vanish. It just shifted directly into the pull request queue.

Senior engineers who once managed a predictable flow of pull requests are now buried under massive backlogs of machine-generated code. Because adding more human reviewers only treats the symptom, teams need to address the friction upstream before code ever gets generated:

Make specifications explicit from the start: Human engineers historically filled in vague requirements using informal judgment and chat threads. AI agents build exactly what is written, meaning ambiguous specs lead to wasted iterations and faulty code.

Involve non-engineers before implementation begins: Pulling product, design, and security into the review stage made sense when writing code was slow. When code is generated quickly, non-technical stakeholders must help shape definitions before generation starts.

Optimize for the cost of late changes: Stop measuring raw pull request volume or review throughput. The metric that truly matters is how expensive it is to redirect work that started from an unclear or incomplete spec.

When you solve requirement ambiguity upstream, you protect your senior reviewers and ensure that fast code generation turns into shipped software rather than review gridlock.

Read the full article:

upsun.com

Beyond pull request counts: why delivery metrics matter for AI workflows ⚡

Flora Brandão — Tue, 27 Oct 2026 15:59:45 +0000

Most engineering teams struggle with a common grind when rolling out AI coding tools: proving that these assistants actually make delivery better. It is easy to watch pull request volume spike, but shipping more code faster is useless if that code simply breaks downstream.

The core challenge is that many organizations never tracked baseline metrics like review turnaround or cycle time before adopting assistants. When teams treat high PR volume as an isolated win, the operational cost often shifts directly into longer review backlogs, unresolved bugs, and support tickets.

Here is what platform teams need to focus on instead:

Measure proven outcomes over raw output: Replace arbitrary productivity guesses with grounded frameworks like DORA. Track deployment frequency, change failure rate, and mean time to recovery to evaluate actual delivery performance.

Remove downstream review debt: Keep deployment pipelines consistent so human reviewers spend time evaluating business logic rather than diagnosing configuration mismatches.

Protect the delivery path: Focus on safe, predictable deployment patterns so automated code changes do not turn into production incidents.

When your delivery pipeline is predictable, your team spends less time fixing broken releases and more time solving meaningful engineering problems.

Read the full article:

upsun.com

Why your local AI agent is a security blindspot 🛡️

Flora Brandão — Tue, 27 Oct 2026 10:00:15 +0000

Running AI agents on local developer laptops creates the lethal trifecta: access to untrusted data, ability to trigger side effects, and visibility into sensitive credentials. Running locally leaves no audit logs, making activity invisible when things go wrong.

The fix requires prioritizing system guarantees over simple guardrails:

Ephemeral sandboxes that reset after each run
Strictly scoped credentials and read-only tokens
Observability and human approval for actions with permanent impact

Building structural controls contains the blast radius before misbehavior propagates to your infrastructure.

Check out the full write-up to dive deeper into building safe agent environments:

Agent security starts with where the agent runs, not how it behaves

Most engineering teams evaluate agent security by asking how the model behaves. Patrick Dawkins, Principal Software Engineer at Upsun, argues that it is the wrong question. The risk starts with where the agent runs and what it can reach from there.

upsun.com

Data residency vs Data sovereignty: What your cloud region actually means ☁️

Flora Brandão — Mon, 26 Oct 2026 15:26:41 +0000

Choosing an EU cloud region feels like an easy compliance win, but hosting data in Frankfurt does not automatically protect you from foreign legal demands like the US CLOUD Act.

Here is what is changing for cloud architecture:

Data residency keeps data inside a boundary, while data sovereignty controls which jurisdiction has legal authority over it.
Regulators expect you to prove your deployed state through APIs, not just point to static policy documents.
Compliance is moving from a global infrastructure setting to a per-workload and per-customer decision.

Check out the full article to see how regulatory shifts in the EU are changing how we choose and audit cloud regions:

Data residency vs. data sovereignty in 2026 | Upsun

Data residency and data sovereignty are not the same thing. Regulators now expect proof of the difference, and IT leaders need to be ready to show it

upsun.com

Managing session storage tradeoffs across your database stack 💾

Flora Brandão — Mon, 26 Oct 2026 10:05:51 +0000

Session storage gets complicated faster than you think. Managing tradeoffs across files, PostgreSQL, Redis, and Valkey often leads to unexpected headaches with expiration, cleanup, and memory limits.

Here is how to navigate session storage:

The problem: Expiration, cleanup, and memory limits can quickly spiral out of control across different datastores.
The fix: Evaluate options like keeping session tables, deleting an hour at a time, or letting the store handle expiration directly.
Why it matters: Testing options on Upsun Cloud helps you optimize memory usage and avoid keeping unnecessary session data around.

Check out the full technical write-up to see how to test session storage tradeoffs with your application data:

Session storage gets complicated faster than you think - Upsun Developer

Explore session storage tradeoffs across files, PostgreSQL, Redis, and Valkey, from expiration and cleanup to memory limits. Test your options on Upsun Cloud.

developer.upsun.com

Fixing the AI bottleneck in software engineering ⚡

Flora Brandão — Fri, 23 Oct 2026 15:00:18 +0000

AI coding tools help developers write code 55% faster, but generating more code does not mean shipping a better product. When AI tools stay trapped on individual laptops, context and custom setups vanish when engineers leave, creating coordination bottlenecks instead of team productivity.

Here is how to fix the development lifecycle:

Validate intent before generating code so you do not build the wrong thing faster
Move context, prompts, and agent setups into a shared team layer
Use traceable workflows and preview environments to verify changes under real conditions

Why it matters:

Treating the full engineering lifecycle as a connected system turns isolated developer speed into real team output.

Check out the full article to dive deeper into closing the AI collaboration gap:

What the full development lifecycle needs | Upsun

AI made writing code fast. See what the rest of the development lifecycle needs: clear intent, shared context, real review, and previews that prove it works

upsun.com

⚙️ How Upsun Dispatch runs workflows from issue to reviewed code

Flora Brandão — Fri, 23 Oct 2026 09:03:29 +0000

Context switching between task managers and pull requests burns developer time. Moving a ticket to reviewed code usually requires a lot of manual setup and manual coordination.

Here is how automated workflow primitives handle the grind:

Integrations translate events from Jira, GitHub, or Linear into a unified format.
Deterministic steps fetch ticket context and handle repository updates safely.
Isolated sandboxes execute planning, coding, and code critiques in temporary environments.
Human gates pause execution so developers can approve plans before code changes start.

This structure removes repetitive setup work while keeping full visibility over every run and context bundle.

Check out the full article to see how these primitives fit together:

Upsun Dispatch: How workflows go from issue to reviewed code

Explore how Upsun Dispatch uses sandboxed agents, tool integrations, and human approval gates to turn Jira issues into reviewed pull requests. Try it today.

upsun.com

Stop assembling audit evidence by hand 🛠️

Flora Brandão — Thu, 22 Oct 2026 15:05:27 +0000

Spending the week before an audit pulling scattered logs out of multiple tools is a manual grind that does not actually make your systems safer.

The problem and fix:

Compliance evidence is usually scattered across CI systems, Git records, and access logs.
Drift happens when one-off configuration changes occur and records fall out of sync.
Automating evidence generation directly at the platform layer turns audit prep from a weeks-long research task into a simple lookup.
Pairing automated deployment logs with Git approval workflows gives auditors the change management proof they need.

Continuous evidence keeps releases from getting blocked by slow, manual verification steps. Check out the full post to read the complete technical breakdown:

Automated audit evidence generation | Upsun

Upsun Cloud logs every deploy, code push, and config change, turning SOC 2 and PCI DSS audit evidence into a byproduct of normal platform operation

upsun.com