Origin load balancing with automatic failover to a standby server
While your site runs on a single server, its outage is the whole site’s outage, and switching to a second one usually happens by hand, after the first complaints. Klacos spreads traffic across your servers, checks continuously that they’re answering, and fails over to a standby server on its own. A server that recovers takes back its share without you lifting a finger.
One server, one point of failure
Adding a second server isn’t enough on its own: someone, or something, has to decide where each visitor goes, notice that a server has stopped answering and stop sending people to it. Done by hand, that happens late, often at night, and in the meantime some of your visitors hit an error.
The standby server you’re paying for then sits idle at the very moment it should be earning its keep. What you want from load balancing: it spots an unhealthy server by itself, stops sending visitors to it before they see an error, and brings it back once it recovers.
When a server falters, the next one takes over
A server that stops answering is set aside: the next visitors go to the others, and a request that can safely be retried is retried on a healthy server. If the whole pool goes down, the standby server takes over. And if nothing answers at all, service continuity serves the last copy of your pages and tells visitors quickly, rather than leaving them waiting.
For a single-page application, a URL the server doesn’t recognise can be sent back to the main page, so the application’s own routing takes over.
When it’s a rush of visitors rather than an outage that threatens your servers, the virtual waiting room queues the overflow in order. And because bot management stops unwanted bots before they get that far, your servers share the traffic that matters, not the scrapers’.
Several servers, one site
Weighted balancing
Health checks
Standby server
By path
Load balancing, service continuity and the waiting room are all part of your site’s performance and availability.
Questions about load balancing
Do my servers need to be in the same place?
No. They’re your origin servers, wherever they are, reachable over HTTP or HTTPS. Klacos spreads traffic across them; it doesn’t replace a network spread over several continents.
What does a visitor see during a failover?
Usually nothing: the unhealthy server is set aside and the next requests go elsewhere. A request that can safely be retried is retried on a healthy server.
Can I take a server out for maintenance?
Yes. You remove it from the pool while you work on it; the others absorb the traffic, then you add it back.
Am I told when a server goes down?
Yes. An alert shows in the console as soon as an origin stops answering.
Stop hearing about server outages from your customers
Request early access, or tell us about your servers and how you handle an outage today.