Degraded is not down
ContenidoContents
When someone in Spain cannot open the site, the question is not whether it is down. It is whether it is down, or whether they cannot reach it.
/status/ is there for that. So I do not argue from memory. And so I do not paint the server as off when the path to it is what failed.
That already happened with LaLiga and the classroom. The origin was answering. From Spain, it was not. This is how I draw it.
Why not Statuspage
Statuspage is for a company to tell its customers: we are down. It charges for that. It looks at the site from the internet, the way a visitor does.
During a match, that look sees a page that will not load. It would paint everything red. On the server, it would still be answering.
I do not want to pay to be lied to. I already have the Cloudflare tunnel, the history and the page. My own PHP, in Spanish and English. The odd case (server fine, Spain not) is not what a rented dashboard solves.
One bar per site
There are six: the site, the blog, the docs, the classroom, the portal and One More.
Each bar is the last 72 hours, Madrid time, one slice per hour. It ends at the current hour. There are no slices in the future. Anything outside those 72 hours is not shown.
The current state is written every five minutes. If that data is older than fifteen, I do not say it is up. I say I do not know.
How I tell an outage
Every five minutes I request each site the way someone outside would.
If it takes too long (five seconds; in practice it dies around thirty), I do not call an outage. I request the same thing on the server.
If the server answers and the outside request does not: that is not an outage. It is that you cannot get there. It goes amber: degraded reachability. No alert. It does not take time off the uptime.
If both fail, then yes: down. That does count.
If the line drops and I cannot even write the check, I log it by hand. Those incidents paint the sites I list and they count too.
When the origin really does not answer, the edge serves “No disponible” with a 503. That is the server off. /status/ is the other question.
The same server
The six bars look like six systems. They are not. They live on the same origin, behind the same tunnel, on the same line.
If the server goes down, they all go down. If the line drops, the same. An ISP block hits the IPs they share: the site, the blog and the docs on one side; the classroom, the portal and One More on the other. The bars are drawn apart. The fate is not.
The check lives there too. If the server is off, it cannot record the outage. That is why I sometimes write it by hand.
And /status/ is on the same site. When the origin really does not answer, you do not get the bars. You get the edge “No disponible”.
That is fine. This is a personal site. I am not selling uptime. I do not need to pretend the classroom and the blog are two different clouds. I need not to lie to myself when Spain cannot get in and the server is still there.
How a site is added
Three places, the same name.
- The service list: what it is called on the page, in Spanish and English.
- The five-minute check: whether it answers right now.
- The timing check: whether it can be reached from outside, or only the server answers.
The name cannot be just the domain if the site is a path on that domain. The blog and the docs live on sergiocomeron.com, like the main site. If I name them by the domain, they get painted on the same bar.
What does not count as down
The weekly classroom maintenance (Monday at 07:00, an hour and a half, a Moodle update) only paints that bar. Blue. It does not count.
A scheduled stop of the server, same rule.
Degraded does not lower the percentage. Uptime is the origin outages plus the incidents I log by hand, if they are not maintenance. Reachability blocks are listed separately.
If there are no outages in the window, the bar shows 100%. I am not making that up: it is a window with no recorded outages.
The page is at sergiocomeron.com/status/. If Spain cannot get in and the server answers, it is not down. It is degraded.