Dev.to Security πŸ” Cybersecurity πŸ‘ 0 πŸ“– 5 min read

Patch week: how to roll a pile of security updates across your sites without breaking them

Every so often the updates stop arriving one at a time. A critical fix for the language runtime lands, the web server ships an advisory, the CMS puts out a security release, and half the plugins on every site push update

Every so often the updates stop arriving one at a time. A critical fix for the language runtime lands, the web server ships an advisory, the CMS puts out a security release, and half the plugins on every site push updates in the same few days. If you look after one site, that's an afternoon. If you look after twenty or two hundred, it's a week that quietly eats everything else.

Across the communities we follow, the same week sounds very different depending on who's living it. An agency counting dozens of fixes and installs in a matter of days and leaning on a half-automated pipeline just to keep up. Developers asking what they should prepare before a critical runtime patch even ships, so they aren't scrambling when it does. People who maintain sites for a living arguing that the only numbers that prove the work is real are things like "restores tested" rather than "updates clicked". Someone who updated a dozen plugins at once, watched something break and had to work out which one did it. And a site owner who tightened their edge firewall rules during all this and found the bots got worse, not better.

We look after other people's WordPress sites, and patch weeks are where good habits either pay off or don't. Here's how we think about it.

The real problem is order, not effort

Most teams will eventually apply every update. The risk lives in the gap between "a fix exists" and "the fix is on this site", and in what breaks when you rush that gap. So the first job isn't updating anything. It's sorting.

We use three lanes.

Lane 1: same day

Things that are exposed and exploitable from the outside, right now:

  • A fix for something reachable without logging in (remote code execution, authentication bypass, file upload flaws).
  • Anything already being exploited, or with a public proof of concept.
  • Components that face the internet directly: the web server, the runtime, login and form handling, anything that accepts uploads.

Lane 1 skips the usual queue. You still take a backup first, but you don't wait for the weekly window.

Lane 2: the next maintenance window

Real fixes with a smaller blast radius:

  • Issues that need a logged-in user, especially a high-privilege one.
  • Plugins that are installed but only lightly used on that site.
  • Security releases of the CMS core that aren't flagged as actively exploited.

These go through your normal staging check and roll out in the scheduled maintenance window.

Lane 3: the routine cycle

Feature updates, cosmetic fixes, and anything with no security note. Patch week is exactly when not to sneak these in, because every extra change makes it harder to tell what broke what.

A patch-week runbook

This is the order of work we'd suggest, whether you run one site or a portfolio.

  1. Know what you have. A simple list of sites with their runtime version, web server, hosting type (managed or not) and the plugins on each. If you can't answer "which of our sites run this plugin?" in a minute, start there. It's the single biggest time-saver in a busy week.

  2. Decide who owns which layer. On managed hosting, the host usually patches the runtime and web server. On a self-managed server, that's you. Many teams discover during a critical runtime advisory that nobody was clearly responsible for that layer. Write it down per site.

  3. Back up, and know the restore works. A backup you've never restored is a hope, not a plan. The people we've seen get through patch weeks calmly are the ones who test a restore routinely, not the ones with the most backups.

  4. Let staging run ahead. Updating a staging copy a day before production catches most of the "this update breaks the checkout" surprises. For a portfolio, even a handful of representative staging sites (one shop, one membership site, one brochure site) catches a lot.

  5. Patch in small batches. If you update twelve plugins at once and something breaks, you now have twelve suspects. Smaller batches, or splitting a failed batch in half until you find the culprit, beats rolling everything back and starting over.

  6. Use the edge as a shield, not a fix. Firewall and rate-limit rules can buy time while you patch: limiting login, form and API endpoints, or blocking a known exploit pattern. But edge rules change behaviour in ways you don't always expect. One common surprise: junk URLs that used to redirect start returning cached errors, and bots hammer them harder. Change one rule at a time and watch your server load afterwards.

  7. Check what actually matters after each batch. Not just "the homepage loads". Can someone log in, submit a form, pay, receive an email? Do scheduled tasks still run?

  8. Keep a short log. Which site, what changed, when, and who checked it. When a client asks "were we affected?", the log is the answer.

Where patch weeks usually go wrong

  • Treating every update as equally urgent. Teams burn out applying cosmetic updates with the same urgency as an exploited flaw, and the important one gets done last.
  • Forgetting the layers under the CMS. "Update all" in a dashboard doesn't touch the runtime, the web server or the operating system. That's where some of the most serious fixes live.
  • Mixing patches with other changes. A redesign, a host move or a new caching layer in the same week makes every breakage ambiguous.
  • Silent clients. Owners don't need technical detail, but a two-line note ("a security fix was applied, here's what we checked") builds more trust than any monthly report.
  • Automation without a brake. Automated pipelines are a real help at scale, and plenty of agencies rely on them. They work best with a pause step for anything touching checkout, login or memberships, and an easy way to hold back a single site.

The trade-offs nobody fully resolves

Auto-updates versus control is the oldest argument here, and both sides are right sometimes. Auto-updates close windows fast and fail loudly. Manual review is safer for complex sites and slower everywhere else. Most teams we know land on a mix: auto-update low-risk sites and minor releases, review anything that handles money or member data.

There's also the cost question. Patch weeks are unbilled work for many maintenance arrangements, and they're the clearest argument for a care plan that names a response time for urgent security fixes, rather than one that only promises "regular updates".

Your turn

We'd genuinely like to compare notes:

  • How do you decide what gets patched the same day versus in the next window? Do you have a written rule, or is it judgment each time?
  • If you look after many sites, what's the one piece of tooling or habit that made a heavy patch week manageable?
  • Have edge or firewall rule changes ever made things worse for you mid-incident? What did you learn?
  • How do you tell clients about security updates without either alarming them or making the work invisible?
πŸ“° Read the original article on Dev.to Security

Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes β€” full credit and traffic to the original publisher.