Page could not be loaded

Bob Noel

Bob Noel

Taxi to Parking
PoA Supporter
Joined
Jun 7, 2008
Messages
27,381
Display Name
Bob Noel
Home Airfield
KBED

Attachments

  • page could not be loaded.jpg
    page could not be loaded.jpg
    94.2 KB · Views: 9
AWS is having problems which ripple everywhere.
 
AWS has melted down today; I don’t know if that’s relevant to this problem, but I know it has slowed my work down dramatically. Like nearly stopped it entirely.
 
A quick & dirty lesson about the Internet. DNS stands for Domain Name Server. IP is the acronym for Internet Protocol. Your IP address may change but your DNS is yours forever, or as long as you pay for it. For example, I have a domain name of my airplane N number but it doesn't exist other than I pay for it, because it's not associated with an IP address at this time.

Best analogy is your cell phone. The number is yours but you can change carriers and keep the number thanks to a Federal law a few years ago. Same with DNS (it's your's) but the IP address can change. There are over 4 Billion unique IP addresses (under IPv4 or 256^4) and we've pretty much run out of them (half belong to the US Government so there's really only 2 Billion) so a few years ago an expanded convention, IPv6, ( or 256^6) was established with over 281,474,976,710,656 IP addresses (the Feds still get half or 281 x 10^12 addresses IIRC or I may be wrong, usually am)

Originally, back in the Dark Ages, there were 7 Authoritative Name Servers, all in the US because we invented this stuff (ANS is the Mount Olympus of DNS/IP) and a hierarchy of lower-level name servers (ANS to Top-level domains or TLD, then Root Names then at the bottom, Resolvers). Don't get me started on port numbers - there are up to 65,000 ports for each unique IP address.
Now, there are at least 13 general purpose ANS, and some single-purpose ANS, such as AWS.

If interested, you're welcome to come to Denver next week when I start teaching this to my beginning computer science students - How the Internet Works. Add to the confusion, servers come and go. To quote Carl Sagan "billions and billions".

This is why it can take up to 24-48 hours to a DNS/IP change to migrate thru the entire system which is no longer just the US but the entire world.

So the problem is the enormous quantity of IP addresses and computers and switches and routers and stuff.

So, when AWS, which is an ANS, craps out, all the requests have to be routed over to other ANS which takes time, when you consider the millions (billions?) of requests every hour. Why so many requests? Every request for a web page, for email, social media, for file transfer, for streaming a movie, for listening to Spotify, and so on and so on ... yada yada yada.

Let me add that I'm not a strong network person, altho I had to study network theory in grad school. This is about as detailed as I get in class. I don't go into the many TCP & UDF packets and the purpose of each and every bit in the various packet headers.

You had to ask, didn't you?
 
Last edited:
Is all that a pedantic way of saying that Amazon can shut down the world? ;)

Looks like the AWS shutdown has more impact than the .gov shutdown.
 
this has nothing to do with aws
 
Half Fast said:
I, for one, will be glad when this damn digital fad ends and we return to analog electronics, like God intended.
Click to expand...
And the Dodgers return to Brooklyn. All will be right with the world.
 
Queensland


Exploding Head Syndrome… It's more common than you think ...
Exploding Head Syndrome… It's more common than you think ...
 
Bob Noel said:
this has nothing to do with aws
Click to expand...
Probably was. Both the POA computing platform and our nameservice live in the AWS cloud. There's a lot of cached stuff and javascript powering the forum software so you can have one of our pages loaded and get issues if the path to the server (or its nameservice) goes away.
 
murphey said:
A quick & dirty lesson about the Internet. DNS stands for Domain Name Server. IP is the acronym for Internet Protocol. Your IP address may change but your DNS is yours forever, or as long as you pay for it. For example, I have a domain name of my airplane N number but it doesn't exist other than I pay for it, because it's not associated with an IP address at this time.

Best analogy is your cell phone. The number is yours but you can change carriers and keep the number thanks to a Federal law a few years ago. Same with DNS (it's your's) but the IP address can change. There are over 4 Billion unique IP addresses (under IPv4 or 256^4) and we've pretty much run out of them (half belong to the US Government so there's really only 2 Billion) so a few years ago an expanded convention, IPv6, ( or 256^6) was established with over 281,474,976,710,656 IP addresses (the Feds still get half or 281 x 10^12 addresses IIRC or I may be wrong, usually am)

Originally, back in the Dark Ages, there were 7 Authoritative Name Servers, all in the US because we invented this stuff (ANS is the Mount Olympus of DNS/IP) and a hierarchy of lower-level name servers (ANS to Top-level domains or TLD, then Root Names then at the bottom, Resolvers). Don't get me started on port numbers - there are up to 65,000 ports for each unique IP address.
Now, there are at least 13 general purpose ANS, and some single-purpose ANS, such as AWS.

If interested, you're welcome to come to Denver next week when I start teaching this to my beginning computer science students - How the Internet Works. Add to the confusion, servers come and go. To quote Carl Sagan "billions and billions".

This is why it can take up to 24-48 hours to a DNS/IP change to migrate thru the entire system which is no longer just the US but the entire world.

So the problem is the enormous quantity of IP addresses and computers and switches and routers and stuff.

So, when AWS, which is an ANS, craps out, all the requests have to be routed over to other ANS which takes time, when you consider the millions (billions?) of requests every hour. Why so many requests? Every request for a web page, for email, social media, for file transfer, for streaming a movie, for listening to Spotify, and so on and so on ... yada yada yada.

Let me add that I'm not a strong network person, altho I had to study network theory in grad school. This is about as detailed as I get in class. I don't go into the many TCP & UDF packets and the purpose of each and every bit in the various packet headers.

You had to ask, didn't you?
Click to expand...
It's a good enough explanation to get a general picture, but for technical accuracy, you seem to be conflating DNS with domain names and authoritative name servers with root servers (I'm unfamiliar with the "ANS' abbreviation). And the reason changes take time to migrate is because of TTLs (time to live)--which used to be set on the order of many hours to days when the Internet was young, but now are generally set much shorter--and caching. And I don't know the details of the AWS outage, but I doubt the repercussions were due to the time needed to reach other root nameservers (milliseconds), since most queries would never go to the root anyway. But who really knows.
 
flyingron said:
Probably was. Both the POA computing platform and our nameservice live in the AWS cloud. There's a lot of cached stuff and javascript powering the forum software so you can have one of our pages loaded and get issues if the path to the server (or its nameservice) goes away.
Click to expand...

nope.

same message for that thread now, just like before the AWS problems ever since that thread was first posted... unless you want to claim the AWS cloud is capable of time travel.
 
Lindberg said:
It's a good enough explanation to get a general picture, but for technical accuracy, you seem to be conflating DNS with domain names and authoritative name servers with root servers (I'm unfamiliar with the "ANS' abbreviation). And the reason changes take time to migrate is because of TTLs (time to live)--which used to be set on the order of many hours to days when the Internet was young, but now are generally set much shorter--and caching. And I don't know the details of the AWS outage, but I doubt the repercussions were due to the time needed to reach other root nameservers (milliseconds), since most queries would never go to the root anyway. But who really knows.
Click to expand...
I just get tired of writing Authoritative Name Servers....hence ANS. The hierarchy is the ANS at the top then the Top-Level domains then the root servers, if my network textbook is still accurate. No doubt the TTLs are down to milliseconds but when you consider the exponentially growing number of root servers, it still takes time. I'm beginning to wonder if some of the records (A and MX) got corrupted. But as I explained, my network background is limited to grad school and fixing the hosts & networks files. There will be the public explanation and somewhere, the real details will come out but the unwashed masses will not know about them. Or care.

I assume most of the readers here are not familiar with the gory details of the networks. Sorta like my beginning students.
 
Bob Noel said:
nope.

same message for that thread now, just like before the AWS problems ever since that thread was first posted... unless you want to claim the AWS cloud is capable of time travel.
Click to expand...
Working fine for me now as it did before.
 
flyingron said:
Working fine for me now as it did before.
Click to expand...

I'm quite sure it is working fine for almost everyone. Going back to my original post, I'm wondering what would cause that error to be thrown for that thread (as example*) and some firefox profile preference setting.

*I've seen it in the past on at least one other thread, I just can't remember which one.
 
murphey said:
The hierarchy is the ANS at the top then the Top-Level domains then the root servers, if my network textbook is still accurate.
Click to expand...
It's either inaccurate or upside down. :dunno:
Here's a decent description of the system: https://bluecatnetworks.com/glossary/what-is-a-dns-server/
murphey said:
No doubt the TTLs are down to milliseconds but when you consider the exponentially growing number of root servers, it still takes time.
Click to expand...
There are still only 13 logical root servers. There don't need to be many because they only point the way to the TLD and ccTLD nameservers.
murphey said:
I'm beginning to wonder if some of the records (A and MX) got corrupted.
Click to expand...
Zone file corruption is a more plausible explanation, but that would have nothing to do with the root servers.
 
It's actually kind of amazing that all of this actually works most of the time. Some very smart people put it all together, with zero idea that it would take off the way it did, let alone scale to the size it has. IP networks, I mean. In the old days, some people would give out not just their email address but the IP address of their mail server, because DNS was new and not always configured everywhere.

Like anything else, some things just crash once in a while. Enough different systems depend on things hosted on AWS that when they don't work right, other things don't work right, either. That would be the same if something went nuts with one of the other two major cloud providers, Google and MS. But overall, I think the reliability picture is a bit better than however many years ago, when all or most of the traffic on the eastern half of the US went through a parking garage or some such in Virginia or Maryland.

Probably way too much stuff is Internet dependent that shouldn't be. For commerce? Sure. The thing that it originally wasn't supposed to be used for is what it's best at. But having it work for mission critical health and safety things, not such a good plan without a backup. Just my ramblings.
 
You must log in or register to reply here.
Back
Top