The week of October 21, 2026, was a rough one for anyone looking after websites. Two critical PHP flaws, nicknamed php2shell, went public and sent hosts racing to patch servers before attackers could use them. We covered how we handled that in our php2shell post.
As if that wasn't enough, in the same week a power outage at a Google Cloud data center in London knocked sites offline, including some that members of our Agency Success community manage.
Neither problem was something a site owner could fix from a WordPress dashboard, and that's exactly why a question about having a plan B came up on Office Hours.
The question: should agencies keep a backup host ready?
One of our viewers runs an agency in the UK and had a handful of client sites go down during the London outage. It was their second force majeure event at that data center, after a fire back in 2018.
They handled it well, sending clients regular updates throughout the day. Because the outage made headline news, the clients understood what was going on and were pleased with how our viewer kept them informed.
During the October 22, 2026 Office Hours livestream, they asked Nathan Ingram, the agency coach at hosting.com, a bigger question:
Should we be considering a throw up hosting solution for future force majeure events to minimise downtime? Or, due to the number of sites and time frame to get re-hosted and reconfigure DNS it's not worth the effort?
In other words, should you have a backup host to put your sites on when the first one goes dark? Nathan's short answer was that you totally could, but for most agencies it won't be worth it. The longer answer, which we'll explore in this post, explains why, and what to do instead.
What does real redundancy involve?
When people picture a backup host, they often imagine a copy of the site sitting somewhere else, ready to go live if needed. True redundancy is a lot more involved than that, and it's worth understanding the moving parts before you price it for a client. Nathan broke it down into a few pieces.
Two servers that stay in sync
A standby copy of a site sounds great at first glance, but then you realize it has to match the live one. For a small site that rarely changes, like a brochure site, using a copy from a week ago might be fine.
For anything with orders, form entries, bookings, new content, or anything to do with new database data, both servers need to have identical copies of the site at all times. Otherwise, when the backup host kicks in it will send visitors to an older version of the site. Anything that happened since the last sync will be missing.
You need tools built for the job to keep two servers matched around the clock, and someone has to keep an eye on them.
Something in the middle directing traffic
Our viewer's question imagined re-hosting the sites and reconfiguring DNS when an outage hits. That's a recovery plan, and a slow one, because DNS changes take time to reach everyone. You also have to do it for every website while the outage is still unfolding.
Real failover works differently. With true redundancy, the site switches over on its own, and that takes a layer sitting between visitors and your servers. It checks whether the main server is responding and sends visitors to the backup automatically. There's no need for anyone to touch DNS records.
If you are not sure where a client's DNS is managed today, our guide on how to find out who handles your website's DNS is a good place to start.
Why it's complicated and costly
Put those pieces together and you're running two hosting environments for one site, plus the tools that keep them in sync and route traffic between them. Someone has to test that the switch works, because a failover that's never been tested is just a hope.
Setting all of this up across different data centers, or even different hosting providers, is possible, but it takes time and it multiplies the hosting bill.
Some of this redundancy already exists on a smaller scale. Cloud hosting spreads a site across several connected servers, so if one machine fails, another takes over. It's one of the differences we cover in VPS vs. Cloud hosting: how to choose the right fit.
A setup like that protects you from a single server failing. However, surviving an outage like the one our viewer described, where the whole data center goes dark, means keeping copies of the site in a different location. That's where things get complicated. Nathan summed up the whole setup like this:
If you have true redundancy, it would just flip over. But you have to have something in the middle that's directing that traffic. And the two servers have to be synchronized with data. It's just complicated. It's very possible. People do it all the time. It's just complicated and costly.
Doesn't Cloudflare already do this?
During the livestream another viewer asked in the chat whether Cloudflare already covers this. It's a fair question, since Cloudflare sits in front of many sites and keeps serving cached content.
However, it only goes so far. Cloudflare caches assets at the edge, like images, scripts, and stylesheets. For many sites, a lot of requests still have to reach the server to get a response. If the server is down, those requests fail.
Cloudflare's Always Online feature helps a little here. When the site can't be reached, it shows visitors a saved copy of the page from the Internet Archive's Wayback Machine. That's better than an error page, but it's a snapshot, not a working website. That means forms won't submit, carts won't check out, and nothing new will show up.
If you want to know more about what Cloudflare does and doesn't do in front of your host, we covered it in Can you use Cloudflare alongside your existing host?
Which client sites need such redundancy?
To determine if you should consider a complex and expensive setup to achieve true redundancy, Nathan's test is simple: what really happens if the site is down for a few hours?
For most sites, the honest answer is “it’s annoying, but nothing terrible happens.” His agency doesn’t mirror any client sites. The closest they have to a critical site is a county commission website, and even that, as he pointed out, isn’t the sheriff’s department or emergency services. It’s important, but the world won't end if it’s offline for an afternoon.
It also helps to think about how often outages like this happen and what they look like when they do. Jumping straight to the biggest solution is perfectly human, but the reality is data center failures are rare. When they do happen, they tend to be big, public events, like our viewer described earlier, and they will be in the news. That's a very different client conversation from one where only their site is down.
And if you have a client that does genuinely run something that's mission-critical, where downtime puts people at risk or costs serious money by the minute, redundancy is worth exploring. That's usually a sign that the site has outgrown conventional hosting, too, and our guide on what enterprise hosting is and how to tell if you need it is a good read.
Let the price make the decision
You don't have to decide on your clients' behalf whether redundancy is worth it. If a client asks for always-on hosting, Nathan suggests treating it as a great conversation to have and then putting a real number on it. "A lot of times that conversation ends when you put a dollar figure on it," he said. Clients often like the idea far more than they like the cost.
To explain why, he compared it to something most of us can relate to:
I would love to walk down to my garage and find a red Ferrari. It would be great, but it's not worth the price to me to have that.
Clients often land in the same place with redundancy, and it's a perfectly fine outcome for most of them, as long as they've made the decision with the full picture. A client who chose to forgo redundancy knows what to expect when an outage happens, which makes the day of the outage much easier for everyone.
What to do instead of a full failover setup
For most agencies, the better investment is making sure you can recover quickly and keep clients informed when something goes wrong. None of this needs a second hosting environment, and it covers far more situations than a data center outage.
Here are a few things we recommend having in place:
Off-site backups you control and a restore you've tested: Your host's backups are the first line, and a second copy stored somewhere else means you're never relying on a single provider. Our article on backing up your data covers the basics.
Uptime monitoring: You want to find out a site is down before your client does. We looked at how monitoring fits alongside Cloudflare in Cloudflare's "Under Attack" mode vs. website monitoring.
Knowing where your host posts updates: During an incident, your host's status page tells you what's happening and when it's expected to be fixed. Ours is at status.hosting.com, and it's where we posted progress throughout the php2shell patching.
Knowing what your host commits to: Read your host's uptime commitment so you know what's covered and what isn't. Ours is explained in the uptime commitment and service credits article.
Together, these give you most of the peace of mind clients are looking for, at a fraction of the cost. They also give you the information you need for the most important part of any outage, which is talking to your clients.
The real plan B is keeping clients in the loop
Before he answered our viewer's question, Nathan pointed out what they had already got right, and it's the part he considers the most important:
The issue is not that something went wrong. The issue is what you do next. And for me, the worst thing you can do is fail to communicate.
Our viewer had no backup host when the London data center went dark, and their clients were still happy because they kept them updated from the start. Nathan's agency did the same thing during php2shell, emailing clients before the news reached them to explain that their sites were covered and the agency was watching closely. In both these cases, clients got what they genuinely wanted to know: what was being done about it, and that someone was there to answer their questions.
So, for most agencies, plan B is not a second host. It's solid backups, monitoring, a host you can trust, and a habit of telling clients what's happening before they have to ask. Save true redundancy for the few sites that need it.
And if you have a question of your own for Nathan, whether it's about redundancy, agency work, WordPress, AI, or anything hosting related, bring it to the next Office Hours livestream. They are every Thursday at 2 p.m. ET, where you can get answers live.
FAQ
What’s the difference between backups, redundancy and failover?
A backup is a saved copy of your site that you restore after something goes wrong, so the site stays down until the restore is finished.
Redundancy means a second, live copy of the site is already running somewhere else. Failover is the automatic switch that sends visitors to that second copy when the first one stops responding.
Backups protect your data, while redundancy and failover protect your uptime. That’s why most sites have backups and far fewer have failover.
How long do DNS changes take to work?
It depends on the record’s TTL (time to live). The TTL tells networks how long to cache the old answer before checking again.
With a short TTL, some visitors see the change within minutes. With longer TTLs, it can take hours, and some networks hold onto the old record even longer. If you’re planning a move, lowering the TTL a day or two beforehand helps the switch happen faster.
What should an outage email to clients include?
Keep it short and factual: what’s happening, whether their site is affected, what’s being done and when you’ll update them next. If the outage is in the news, a link to a reliable report helps show it isn’t just them. Most of all, send it early, ideally before the client notices the problem or hears about it somewhere else.
Should you talk to clients about outages before one happens?
Yes, it makes the day of an outage much easier. A short note in your maintenance agreement or onboarding can explain how you handle downtime, how you’ll contact them and what redundancy would cost if they ever want it. Clients who already know the plan are far less likely to panic, and the ones who need more protection can ask for it before they need it.