Quote:
Originally Posted by gmw
From what I've seen the initial load problem (at least) does seem to be DNS related. I've been using CloudFlare for DNS and I've been wondering if I should change, but most other sites have been working okay.
|
What interests me is that this happens on only one of your connected devices. I would triple check the settings on that machine. External DNS resolution problems would bite all of your machines.
There are several third-party DNS resolver services. Google has one. OpenDNS is another. Cloudflare's is a third.
What I use is OpenDNS. Like almost everyone, I have a home network, and devices connect to my WiFi router, (The desktop connects via CAT5 cable directly to the route4. MY SO's laptop and mine connect via WiFi, as do several Android devices. A plus for OpenDNS is that it maintains a badlist of sites known to serve malware and blocks access to them. I looked at Cloudflare's offering, but saw no compelling reason to switch. I am not that concerned by privacy, and Cloudflare was not noticeably faster than OpenDNS
here.
Cloudflare is widely used by sites as a shield against things like DDOS attacks, and lots of folks were interested when they announced a DNS service touting speed and ease of configuration., with special attention to privacy, and it got widely adopted.
But any centralized service can have issues. Last year, the annual World Science Fiction Convention was to take place in Wellington, NZ. COVID-19 happened. In person gatherings were not possible, so CoNZealand decided to go virtual, using Zoom for programming and Discord for social stuff. I found myself on the tech staff assisting in this.
During the run up to the con, staffers suddenly discovered access to various resources had gone away. The first thought was of course "It's a problem on my machine. Recheck all network settings and reboot." That did not solve the problem.
The underlying issue was a configuration error on Cloudflare's end. They have major servers in a number of locations around the world. They were trying to do load balancing, and redirect traffic from a heavily loaded server to others to reduce congestion. Unfortunately, they made an error in the command to do that. Instead of redirecting
from the congested server
to others, it redirected traffic
from the other servers
to the one whose congestion they were trying to reduce. Essentially, they DDOSed themselves. This had all sorts of side effects, like impacting DNS resolution services to folks who used Cloudflare as primary DNS resolver.
(In my setup, primary and secondary DNS resolution is pointed at OpenDNS, but my ISP's DNS server is a worst case fallback If they are
all down, I likely have worse problems than being able to reach the internet.
)
This was one of the more dramatic "Oops!" moments I have seen on the Internet. I thought about it when a configuration issue at Facebook made Facebook, Instagram, and WhatsApp unavailable for a period. This has produced calls to break up Facebook, which also owns Instagram and WhatsApp. Unfortunately, that's not really a solution for the problem.
______
Dennis