When the automation stops, so does the business
We get the call when something has already gone wrong. An invoice did not fire. A lead sat in a form for two days because the workflow that was supposed to notify the sales team never ran. Somewhere in there, n8n uptime and infrastructure monitoring stopped being a technical detail and became the reason a client lost trust. If you have handed your bookings, your billing, or your customer replies over to an automation, uptime is not a nice-to-have. It is the business.
What happens when n8n goes down
Picture a Tuesday. Your chatbot is answering enquiries while you are out on a job. Your CRM is updating itself when a form gets filled in. Your invoices are going out the moment a job is marked complete. Now picture all of that stopping for an hour, or a day, without anyone noticing until a customer rings to ask why nothing happened. That is the real cost of downtime. It is not an abstract percentage on a sales page. It is a missed booking, a late invoice, a customer who assumes you ignored them.
The more you automate, the more exposed you are if the thing running your automation falls over. That is not a reason to automate less. It is a reason to be fussy about what your automation runs on.
n8n uptime and infrastructure monitoring: what the numbers actually mean
Uptime gets quoted as a percentage, which makes it sound tidier than it is. A 99.99 percent uptime commitment works out to roughly 52 minutes of downtime a year. That is a very different experience to a platform that only promises 99 percent, which allows for closer to 87 hours of downtime a year. Fifty two minutes you can plan around. Eighty seven hours, spread unpredictably across the year, you cannot.
This is why we run our hosting on Kubernetes infrastructure with a 99.99 percent uptime SLA, rather than a single server that quietly goes down on a Sunday night with nobody watching. An SLA is only worth something if someone is actually monitoring against it. A number on a website is a promise. Monitoring is the proof.
Why self-hosting without monitoring is a gamble
Self-hosting n8n on your own server has real upside. You get full control of your data, no execution limits, and none of the creeping subscription costs that come with tools like Zapier or Make once your workflows grow past a free tier. If you want the detail on how that comparison actually plays out for a small team, we wrote about it in n8n vs Zapier vs Make: what actually matters for a small team.
But control comes with a job attached. Someone has to watch the server, apply updates, manage backups, and notice when a workflow quietly stops executing. A VPS running n8n needs proper resourcing, and heavier automation or AI-driven workflows need more than a bargain basement box with 2GB of RAM. None of that is difficult. It is just ongoing, and it is easy to let slide when you are busy running the actual business. If you are not comfortable being the person who gets paged at midnight, that is worth being honest with yourself about before you self-host.
Monitoring workflow executions, not just the server
Server uptime is only half the picture. A server can be up and the workflow can still be failing. A third party API changes its response format, a credential expires, a rate limit gets hit, and the workflow quietly stops doing its job while the server itself reports perfectly healthy. This is a real question people ask in the n8n community: how do you actually monitor whether executions are succeeding, not just whether the box is switched on.
The answer is that infrastructure monitoring and execution monitoring are two different jobs, and both need doing. One tells you the engine is running. The other tells you the car is actually moving in the right direction. We watch both for the businesses on our managed plans, because a workflow that fails silently for three weeks is worse than one that fails loudly on day one.
What we do differently
We host n8n on Kubernetes infrastructure with a 99.99 percent uptime SLA, uncapped executions on flat pricing, and no lock-in contracts. That combination matters more than it sounds. Uncapped executions means you are not throttled the month your automation actually starts working well. Flat pricing means a busy quarter does not blow out your invoice. No lock-in means we have to keep earning the account, not just bill it.
Support runs in tiers, from self-managed hosting where you keep control and we keep the infrastructure solid, through to a fully embedded automation team for businesses that want automation handled end to end. We set out the difference between the middle tiers in Support Plus vs Pro: getting the most from self hosted n8n, and the hosting details themselves are on our n8n hosting page. We have done this for over 300 businesses across Australia, New Zealand and beyond, including Australian Writers’ Centre and MCG Quantity Surveyors, and the thing that keeps coming up is the same one we have been talking about here. People do not notice good infrastructure. They only notice bad infrastructure.
What to ask before you trust your automation to anyone
Before you commit your bookings, your billing, or your customer replies to any automation setup, ask what the uptime SLA actually is, ask who monitors it, and ask what happens when a workflow fails quietly rather than loudly. Ask whether you are locked into a contract if the answers turn out to be weak. If the person selling you the platform cannot tell you their own uptime number without checking, that tells you something too.
Good infrastructure is boring on purpose. You should not be thinking about your n8n uptime and infrastructure monitoring at all, because it should just work in the background while you get on with your Tuesday. The moment you start thinking about it is usually the moment it has already let you down.
