Skip to main content

What 1.8 Million Outages Say About Who Is Really On Call in August

Mattias Geniar
Oh Dear

Website outages get resolved just as fast in August as they do in November. I went looking for the opposite: the summer slowdown everyone assumes is there once the people who fix things are away. It isn't in the data we collected, covering 1.8 million confirmed outages across tens of thousands of websites.

Over the last 17 months — two summers, one of them with a completed August — the median time to resolve an outage held at roughly two minutes. That's not one average for the whole period. It's the median for each month on its own, holiday months included. Last August, the emptiest month on most office calendars, was one of the fastest months in the set. Summer as a season came in slightly ahead of the rest of the year, not behind it.

That surprised me, and I suspect it would surprise most people running infrastructure.

The median is doing the work here, and it's worth being clear about why. It's the midpoint outage, the typical one. Monthly averages move for reasons that have nothing to do with seasons: one marathon incident, a certificate that expired on a Friday night and got noticed on Monday, drags a whole month's mean with it. A mean can sit several times above the median in the same dataset.

So if you want to know whether your recovery is degrading in summer, look at the middle of the distribution and at the tail separately. Judge recovery on the typical outage, then go and look at the worst one on purpose. A blended average mostly tells you how unlucky you got.

Why the line stays flat is the more interesting question, and the data doesn't answer that on its own.

Some of it is automation. Platforms restart processes, health checks pull a bad instance out of rotation, a deploy rolls itself back. None of that knows what month it is.

But plenty of fixes still need a person: a config change reverted, a queue drained, a stuck deploy killed by someone who knows which one it is. Somebody opened a laptop. Nothing in the numbers tells me how that split moves between July and November. Only that the result doesn't budge.

My read is that in a large organization, a flat summer is a staffing outcome. You've got follow-the-sun coverage, a documented rotation, someone who's paid to be reachable. Summer cover is a scheduling exercise that somebody owns in advance.

Most of the web isn't that. It's small development teams and agencies where keeping the site up was never formally assigned to anybody. One person has the alerts on their phone and has never been asked whether they mind. I think a good part of that flat August is them, answering.

If it is, the number is worth less than it looks. A resilience figure that holds because one unnamed person never quite logs off isn't in the runbook or the on-call budget, and it doesn't show up on a dashboard. It looks exactly like automation working, right up to the August where that person is genuinely unreachable, or has left.

I spent years as a sysadmin, so my first reaction to this was pride in the people doing the work. What bothers me is the structure they're doing it inside.

There's a straightforward exercise here for anyone with a summer coverage plan. Take last month's incidents and write down who actually resolved each one. Not the person who owned the service on paper. The person who did the work. If the same name keeps appearing and that name isn't on any rotation, you've found the layer your uptime is resting on. That's worth knowing before you find out the hard way, and it's worth saying out loud to the person concerned, ideally alongside some form of thanks more durable than cake.

Mattias Geniar is Co-Founder of Oh Dear

The Latest

Rapid AI adoption and the unique ways AI workloads operate is redefining the scope and structure of what these teams must deliver. This shift is forcing organizations to rethink how they manage scale, automation, and control, according to The State of SRE and Platform Engineering 2026, a new report from Dynatrace ...

AI is usually talked about as a software tool, but it also depends heavily on the network behind it. Whether a company is using AI for chatbots, automation, monitoring, analytics, or employee support, all of that information has to move across the network in a reliable and secure way. That means AI is not just an application decision. It is also an infrastructure decision. Before organizations rush into AI, they should ask a simple question: Is our network ready to support it? ...

Enterprise AI often lacks governed access to where business processes actually execute. Without that access, AI agents may be able to reason, but they cannot operate reliably across enterprise workflows. For AI agents to effectively carry out workflows, they will require integration-layer context and controls. Organizations can implement these prerequisites by providing AI with managed access to the middleware layer ...

Enterprise networks rarely behave the same way for very long. A routing adjustment in one region may unexpectedly alter application performance in another. A cloud migration may introduce hidden dependencies that go unnoticed until an outage occurs. All the while, the network is managed by several different teams, each of whom use different tool sets — and as a result, have different views of the network ... There’s usually an engineer who remembers why traffic fails over a certain way between sites, or which transparent firewall was added where. The problem is that human memory cannot scale alongside enterprise-scale networks ...

Ask an infrastructure team how confident they are in their ability to govern AI, and most will tell you they've got it handled. A recent survey of 406 IT decision-makers and platform engineering leaders found 86% expressing exactly that confidence. Ask the same group whether they have a formal written AI governance policy, and the number drops to 30%, according to Spacelift's Infrastructure Automation Report ...

In MEAN TIME TO INSIGHT Episode 27, Shamus McGillicuddy, EMA VP of Research, Network Infrastructure and Operations, and Parker Hathcock, EMA Research Director covering IT Service/Operations (ServiceOps), discuss observability unification in modern IT operations ... 

Virtual Private Networks became a cornerstone of enterprise security at a time when corporate infrastructure looked very different from today ... For years, this model worked well. But the architecture behind VPNs assumed a centralized corporate environment—one where the network itself was the hub of activity. In a cloud — first world, that assumption no longer holds ...

Website outages get resolved just as fast in August as they do in November. I went looking for the opposite: the summer slowdown everyone assumes is there once the people who fix things are away. It isn't in the data we collected, covering 1.8 million confirmed outages across tens of thousands of websites ...

This year, many of the cloud infrastructure contracts signed in the early days of the AI boom will come up for renewal. As the year goes on, I anticipate we'll see a significant amount of cloud vendor swapouts and multi-cloud adoption, and the reason isn't just GPU depreciation. It's because they're tired of their current cloud providers ...

There's a moment the many observability teams have experienced days into bringing a new service into production: you realize that the vendor's claims of "intelligent" behavior included a large serving of hype. Their dashboards look nice until they don't, the failure modes are a black box, and no one on the team can confidently explain why the system did what it did at 2 am. Agentic AI is about to force every Ops team to relive that moment at web-scale until they start treating these systems as the dependencies they actually are ...

What 1.8 Million Outages Say About Who Is Really On Call in August

Mattias Geniar
Oh Dear

Website outages get resolved just as fast in August as they do in November. I went looking for the opposite: the summer slowdown everyone assumes is there once the people who fix things are away. It isn't in the data we collected, covering 1.8 million confirmed outages across tens of thousands of websites.

Over the last 17 months — two summers, one of them with a completed August — the median time to resolve an outage held at roughly two minutes. That's not one average for the whole period. It's the median for each month on its own, holiday months included. Last August, the emptiest month on most office calendars, was one of the fastest months in the set. Summer as a season came in slightly ahead of the rest of the year, not behind it.

That surprised me, and I suspect it would surprise most people running infrastructure.

The median is doing the work here, and it's worth being clear about why. It's the midpoint outage, the typical one. Monthly averages move for reasons that have nothing to do with seasons: one marathon incident, a certificate that expired on a Friday night and got noticed on Monday, drags a whole month's mean with it. A mean can sit several times above the median in the same dataset.

So if you want to know whether your recovery is degrading in summer, look at the middle of the distribution and at the tail separately. Judge recovery on the typical outage, then go and look at the worst one on purpose. A blended average mostly tells you how unlucky you got.

Why the line stays flat is the more interesting question, and the data doesn't answer that on its own.

Some of it is automation. Platforms restart processes, health checks pull a bad instance out of rotation, a deploy rolls itself back. None of that knows what month it is.

But plenty of fixes still need a person: a config change reverted, a queue drained, a stuck deploy killed by someone who knows which one it is. Somebody opened a laptop. Nothing in the numbers tells me how that split moves between July and November. Only that the result doesn't budge.

My read is that in a large organization, a flat summer is a staffing outcome. You've got follow-the-sun coverage, a documented rotation, someone who's paid to be reachable. Summer cover is a scheduling exercise that somebody owns in advance.

Most of the web isn't that. It's small development teams and agencies where keeping the site up was never formally assigned to anybody. One person has the alerts on their phone and has never been asked whether they mind. I think a good part of that flat August is them, answering.

If it is, the number is worth less than it looks. A resilience figure that holds because one unnamed person never quite logs off isn't in the runbook or the on-call budget, and it doesn't show up on a dashboard. It looks exactly like automation working, right up to the August where that person is genuinely unreachable, or has left.

I spent years as a sysadmin, so my first reaction to this was pride in the people doing the work. What bothers me is the structure they're doing it inside.

There's a straightforward exercise here for anyone with a summer coverage plan. Take last month's incidents and write down who actually resolved each one. Not the person who owned the service on paper. The person who did the work. If the same name keeps appearing and that name isn't on any rotation, you've found the layer your uptime is resting on. That's worth knowing before you find out the hard way, and it's worth saying out loud to the person concerned, ideally alongside some form of thanks more durable than cake.

Mattias Geniar is Co-Founder of Oh Dear

The Latest

Rapid AI adoption and the unique ways AI workloads operate is redefining the scope and structure of what these teams must deliver. This shift is forcing organizations to rethink how they manage scale, automation, and control, according to The State of SRE and Platform Engineering 2026, a new report from Dynatrace ...

AI is usually talked about as a software tool, but it also depends heavily on the network behind it. Whether a company is using AI for chatbots, automation, monitoring, analytics, or employee support, all of that information has to move across the network in a reliable and secure way. That means AI is not just an application decision. It is also an infrastructure decision. Before organizations rush into AI, they should ask a simple question: Is our network ready to support it? ...

Enterprise AI often lacks governed access to where business processes actually execute. Without that access, AI agents may be able to reason, but they cannot operate reliably across enterprise workflows. For AI agents to effectively carry out workflows, they will require integration-layer context and controls. Organizations can implement these prerequisites by providing AI with managed access to the middleware layer ...

Enterprise networks rarely behave the same way for very long. A routing adjustment in one region may unexpectedly alter application performance in another. A cloud migration may introduce hidden dependencies that go unnoticed until an outage occurs. All the while, the network is managed by several different teams, each of whom use different tool sets — and as a result, have different views of the network ... There’s usually an engineer who remembers why traffic fails over a certain way between sites, or which transparent firewall was added where. The problem is that human memory cannot scale alongside enterprise-scale networks ...

Ask an infrastructure team how confident they are in their ability to govern AI, and most will tell you they've got it handled. A recent survey of 406 IT decision-makers and platform engineering leaders found 86% expressing exactly that confidence. Ask the same group whether they have a formal written AI governance policy, and the number drops to 30%, according to Spacelift's Infrastructure Automation Report ...

In MEAN TIME TO INSIGHT Episode 27, Shamus McGillicuddy, EMA VP of Research, Network Infrastructure and Operations, and Parker Hathcock, EMA Research Director covering IT Service/Operations (ServiceOps), discuss observability unification in modern IT operations ... 

Virtual Private Networks became a cornerstone of enterprise security at a time when corporate infrastructure looked very different from today ... For years, this model worked well. But the architecture behind VPNs assumed a centralized corporate environment—one where the network itself was the hub of activity. In a cloud — first world, that assumption no longer holds ...

Website outages get resolved just as fast in August as they do in November. I went looking for the opposite: the summer slowdown everyone assumes is there once the people who fix things are away. It isn't in the data we collected, covering 1.8 million confirmed outages across tens of thousands of websites ...

This year, many of the cloud infrastructure contracts signed in the early days of the AI boom will come up for renewal. As the year goes on, I anticipate we'll see a significant amount of cloud vendor swapouts and multi-cloud adoption, and the reason isn't just GPU depreciation. It's because they're tired of their current cloud providers ...

There's a moment the many observability teams have experienced days into bringing a new service into production: you realize that the vendor's claims of "intelligent" behavior included a large serving of hype. Their dashboards look nice until they don't, the failure modes are a black box, and no one on the team can confidently explain why the system did what it did at 2 am. Agentic AI is about to force every Ops team to relive that moment at web-scale until they start treating these systems as the dependencies they actually are ...