Routing Problems
- Thread starter daft27
- Start date
Ad: Buy Girls Und Panzer Merch from Play Asia!
QUOTE (daft27 @ Jun 03 2012, 01:36 PM) We've been receiving reports of excessive connectivity problems today. To address this problem, we are currently rerouting traffic away from the problematic server. Unfortunately, it may take a few hours for the changes to fully take effect.
I was starting to wonder what was up since I couldn't even access with a proxy. We haven't had this problem for quite some time now...
I was starting to wonder what was up since I couldn't even access with a proxy. We haven't had this problem for quite some time now...
The London Internet exchange also had some serious issues during the week, seems a lot of places are trying to make some upgrades which are not going quite perfectly.
So in other words there was a supposed massive upgrade for the Internet and it failed huh?
Sorry I had to say it. In any case sounds like there is a bad update for the servers though, but I hope it won't cause much of a hassle and things will run smoothly once again.
Sorry I had to say it. In any case sounds like there is a bad update for the servers though, but I hope it won't cause much of a hassle and things will run smoothly once again.
QUOTE (Kid-Wolf @ Jun 04 2012, 02:18 AM) So in other words there was a supposed massive upgrade for the Internet and it failed huh?
Sorry I had to say it. In any case sounds like there is a bad update for the servers though, but I hope it won't cause much of a hassle and things will run smoothly once again.
Its more about some of the devices that connect different networks together, as the the networks and links get more full, they need faster routers and more lines. Sometimes an update can go wrong, or theres a bug in the software.
Routers are constantly sending messages to each other to tell them what networks connect to them, it only takes one message to be wrong and it changes how traffic on those near by move around.
Its not actually anything on our / FTV's servers that cause it.
Sorry I had to say it. In any case sounds like there is a bad update for the servers though, but I hope it won't cause much of a hassle and things will run smoothly once again.
Its more about some of the devices that connect different networks together, as the the networks and links get more full, they need faster routers and more lines. Sometimes an update can go wrong, or theres a bug in the software.
Routers are constantly sending messages to each other to tell them what networks connect to them, it only takes one message to be wrong and it changes how traffic on those near by move around.
Its not actually anything on our / FTV's servers that cause it.
I thought I got banned.
I was able to post in the anime section through the FTV search engine but I can't access it normally.
See if you guys can use this link:-
http://boards.fansub.tv/index.php?showforum=3
Because this is the link I get when I press Anime at the forum's front page:-
Ahh...I couldn't get the "different" link, looks like its been fixed.
See if you guys can use this link:-
http://boards.fansub.tv/index.php?showforum=3
Because this is the link I get when I press Anime at the forum's front page:-
Ahh...I couldn't get the "different" link, looks like its been fixed.
Fansub.TV uses a reverse proxy to filter/cache/pipeline requests to the webserver.
Normally the FTV error page comes up automatically indicating that there's some sort of temporary problem at the reverse-proxy level, but the issue today manifested itself in a way which caused pages (and other image, css, and js assets) to be served and cached incorrectly.
This issue normally occurs when the reverse-proxy caches bad data, but it can becomes even worse when the bad data gets cached in the browser itself. Even after the problem is fixed on the server end, the problem may still persist for various users if their browser still has a copy of the bad data. (You can actually verify this problem by testing with two different browsers, or two different computers on the same network.)
FTV also uses a third party failover service to detect and actively remove disabled servers. Unfortunately, this can also make matters worse if the failover service also has a copy of bad data AND the the user's ISPs cache the failover operation routing instead of the normal operation routing.
To fix a bunch of these problems, we've changed the user-facing infrastructure by getting rid of the failover service and by modifying the configuration of our reverse-proxies. The reverse proxies are now located in 13 different datacenters around the world where users are automatically routed to the nearest geographical location. If the connection goes down, theoretically the reverse-proxy can maintain limited service by displaying a cached version of the content. The new reverse-proxies should also be able to recognize when bad data is being served, and request new content.
New setup:
To summarize, this new setup should be better than the last.
Normally the FTV error page comes up automatically indicating that there's some sort of temporary problem at the reverse-proxy level, but the issue today manifested itself in a way which caused pages (and other image, css, and js assets) to be served and cached incorrectly.
This issue normally occurs when the reverse-proxy caches bad data, but it can becomes even worse when the bad data gets cached in the browser itself. Even after the problem is fixed on the server end, the problem may still persist for various users if their browser still has a copy of the bad data. (You can actually verify this problem by testing with two different browsers, or two different computers on the same network.)
FTV also uses a third party failover service to detect and actively remove disabled servers. Unfortunately, this can also make matters worse if the failover service also has a copy of bad data AND the the user's ISPs cache the failover operation routing instead of the normal operation routing.
To fix a bunch of these problems, we've changed the user-facing infrastructure by getting rid of the failover service and by modifying the configuration of our reverse-proxies. The reverse proxies are now located in 13 different datacenters around the world where users are automatically routed to the nearest geographical location. If the connection goes down, theoretically the reverse-proxy can maintain limited service by displaying a cached version of the content. The new reverse-proxies should also be able to recognize when bad data is being served, and request new content.
New setup:
To summarize, this new setup should be better than the last.
