I just asked Trevor what was going on. Apparently we turned off the restart-HN process back when we were hacked a few days ago, and we forgot to turn it back on. Things should be back to normal now.
It looks like it's just up or down on the y-axis. RRDTool is often used to generate more complex graphs, so something this simple isn't always the most intuitive.
The monitor checks HN once per minute and records a 0 for up or a 1 for down in the database.
RRDTool then averages these values out nicely so that, for example, when the graph is at weekday-resolution and you see a 50% filled bar spanning one day, then HN was down half of that day.
My recollection was that it was generalized so that a user could specify a site, and the app would say whether the site was up or not .. but it did not monitor to let you know when it was back up ..
Combining the two would make a very handy tool .. to be able to monitor when any site is back up ..
Are you referring to HN during the earlier downtime?
The server still seemed to be responding, it was just empty. (I didn't check the headers, so I don't know specifically what if anything was happening.) I believe dfeojm works by checking whether you get a response or not, so as far as it could tell HN was still up.
I am finding that HN is down more often these days. Is it because of occasional spikes or is it a persistent occurrence. Perhaps pg could explain what's going on under the hood?
It's like a global noprocrast. Let's take the next leap and add support for a 'global noprocrast vote'. If more people vote 'yes' than vote 'no', the site goes down for an hour.
I'd say that's a sign you probably should block yourself from HN for a while. Any time that I notice that I'm getting too compulsive about a particular site I add it to a blocked list and leave it there for a week or two. I'm also a huge fan of Freedom.app for disconnecting my internet for blocks of time where I notice that I'm too easily distracted.
Not that I know of, but it's a pretty trivial app. It just removes all of the default routing. Sure, I could probably add them back, but I've intentionally never learned how to use the BSD route that comes with Mac OS for that reason. :-)
With Linux, actually, you could probably go one level lower by writing a small kernel driver that you could signal from user space to disable networking for a given period and then do some goofiness so that it couldn't be unloaded effectively and still get networking back.
It might be useful to add what timezone that time actually refers to. Unless we sometime soon have downtime which spans several days which makes the +-12 margin irrelevant.
Sheesh - it's a free service on the internet, used (amongst other things) as a vehicle for the new language he's working on. I think you're setting your expectations a bit off.
What does the graph really provide? We know when it's down -- Just seems like griping for me.
