Conversation

So, the top half is ping times (to these hosts, about 10-20ms) and the bottom half is packet loss, as a percentage, jumping around wildly.

For the first image, those are some of the machines I operate (including FSE) so I have them on that screen usually to see if they're up. The ping times are normal when loss is low, about 10-20ms, but the packet loss is fluctuating wildly. They're on different networks, so obviously it's probably the local connection if they are *all* like that. (It's unlikely *all* of those machines would get swamped simultaneously.)

The second and third images, I tried pinging (in order) Level 3, Google's DNS, and Cloudfed's DNS, but tweaked the interval from 5s to 250ms, so the timescale is compressed by a factor of 20, and you can see the really weird pattern, no packet loss for a bit, then 100%, then 50%, then back to 0%. (The inline measurements, stuff like "14:32:18 0.0158" is the time and then the ping time in seconds, measured at the time the vertical line indicates. Same thing with below, but for packet loss percentage.)

When it was completely down a couple of days ago, the ISP said that one of the upstream fiberoptic lines broke. Even odds that this is Harmful Misinformation, as a guy tells his boss something and then his boss tells his boss, and then that boss sends the memo to tech support, and then tech support tells the customer, so maybe it was the fiberoptic lines or maybe someone spilled beer into some networking gear or maybe they're setting up horrifying :glowing: government surveillance equipment :glowinthedark: or an almost equally horrifying :shitassdickfuckfluoride: analytics recording device :bezos: or whatever. Probably whatever is happening today is related to whatever was happening before.

I haven't called back today, but if they say it's got to do with the line again, I'd suspect the broken fiberoptic line story is bullshit because today, it's cycling almost exactly every thirty seconds: the bad part lasts 20s, then there's 0% loss for 10s, and that's been going on for about an hour. If it's the line, I don't know how the *line* would know that 30 seconds have passed: it seems like too consistent an interval for that. This seems like a dying router (or an otherwise unreliable piece of gear, like maybe there's a new GFW or they decided to roll out some new ad-targeting analytics infrastructure or something). You get regular intervals from software or circuits: if it's just a line, the line is either consistently up or consistently down or consistently somewhat lossy.

When the net's unreliable, sometimes it's possible to double-post (because the server gets the post but the connection is interrupted before it responds with a confirmation to the browser), but in this case, if I click "Post" when we're inside the ten-second "good" window, it should go through. Whatever the issue actually is, it is kind of convenient that it's happening on a predictable interval. :terrylol2:
spiky_loss.gif
pattern.gif
pattern_persists.gif

Public

Notices

Feeds