POL News: All Worlds Maintenance

[Everlasting]

/poke Art for more info about the corrupted headers, it certainly could be the problem.

If this is the case why aren't we seeing corrupted packets in all other Internet applications? Why is the problem seemingly only isolated to FFXI and nobody else in the world is reporting problems?

 

[Artem]

SE's got a much better shot at figuring it out and fixing it than I do, they just choose not to.

That is the most retarded statement ever. They are working on it for sure. Backbone problems don't get fixed in a day. When the backbone in russia collapsed, it was out for like 3 weeks lol.

 

[Uzor]

The buffs are a DAT mod - that is not SE's problem and people need to expect degraded support when they start hacking programs. The argument being made is that the update and some problem with a backbone is mere coincidence. They happened at the same time independent of each other. While it is possible - I do not see a problem with an internet backbone, being capable of determining how one zone is different from another zone in XI. If 1 zone was broken for everyone, perhaps a keyword in a packet is causing the packet to be dropped - but this is not the case. Particular zones are innaccessable to some people. The game itself is broken for others. The only thing unchanging is which zones are being broken.

I feel something SE has done to prevent DDoS attacks as well as hacking attempts has begun to wreak havoc on the service inadvertently. You can agree or disagree, expand or dismiss my belief - but I just do not understand how an error in 'the internet' can discriminate the functionality of a program which such unerring static nature unless it was programmed to do so.

To answer how it can happen after maintaince ->

If SE replaced some of their network equipment, like gateways etc, which I do believe they do to keep their equipment fresh, things would have a different ethernet address. In this case, your packets need to "learn" all over again where to go, resulting with your packets being sent through a route they didn't previously take.

Now, I don't remember if it was called an ethernet address as it were 4 years ago I took the courses, but I do know that every gateway etc. have a unique address that things get sent to. If I remember right, IP addresses are connected to ethernet addresses. So what things realyl get sent to is the ethernet addresses. Thus, if things get sent to a new ethernet address currently not on the list, they need to fidn them again.

 

[Uzor]

That is the most retarded statement ever. They are working on it for sure. Backbone problems don't get fixed in a day. When the backbone in russia collapsed, it was out for like 3 weeks lol.

Yes T_T You had problems during that time too? ; ;

 

[Arctyrael]

To answer how it can happen after maintaince ->

If SE replaced some of their network equipment, like gateways etc, which I do believe they do to keep their equipment fresh, things would have a different ethernet address. In this case, your packets need to "learn" all over again where to go, resulting with your packets being sent through a route they didn't previously take.

Now, I don't remember if it was called an ethernet address as it were 4 years ago I took the courses, but I do know that every gateway etc. have a unique address that things get sent to. If I remember right, IP addresses are connected to ethernet addresses. So what things realyl get sent to is the ethernet addresses. Thus, if things get sent to a new ethernet address currently not on the list, they need to fidn them again.

Upper Level Domain Name Servers refresh all routing information every 24 hours. Thats the typical timeout on cached server info. If it had to do with routing - it would have reared its ugly head for a day, and gone by the wayside.

 

[Uzor]

Upper Level Domain names server refresh all routing information every 24 hours. Thats the typical timeout on cached server info. If it had to do with routing - it would have reared its ugly head for a day, and gone by the wayside.

Not neccesarily. Like Artem said, when the backbone in Russia crashed, I (and most FFXI players in Sweden) had problems for several weeks. I don't remember when, but one of the times the problem persisted for close to 2 months.

 

[CmdrFly]

Now, I don't remember if it was called an ethernet address as it were 4 years ago I took the courses, but I do know that every gateway etc. have a unique address that things get sent to. If I remember right, IP addresses are connected to ethernet addresses. So what things realyl get sent to is the ethernet addresses. Thus, if things get sent to a new ethernet address currently not on the list, they need to fidn them again.

Ethernet address? Never heard of it, unless you're talking about MACs. Each device does have a unique hostname and DNS suffix, but that ultimately maps back to an IP.

Upper Level Domain names server refresh all routing information every 24 hours. Thats the typical timeout on cached server info. If it had to do with routing - it would have reared its ugly head for a day, and gone by the wayside.

What he said. BGP is smarter than your average human, and a hell of a lot less lazy. If a link goes down, or is changed, routes are re-established in minutes and within a day the new route is propagated to peers.

 

[CmdrFly]

Not neccesarily. Like Artem said, when the backbone in Russia crashed, I (and most FFXI players in Sweden) had problems for several weeks. I don't remember when, but one of the times the problem persisted for close to 2 months.

When was this? Not ringing a bell to me, I'd like to read more on it. If things took 3 weeks to re-establish, it likely wasn't a routing problem unless some noob was running a protocol with misconfigured options.

 

[Artem]

/sigh I didn't really want to go into the technical details about it. The servers aren't like based on the server name... There is *no* quetz server. Quetz serve is actually composed of many servers. They are divided by areas mostly. Each server doesn't just hold on area, they hold a few. Each of these (let's call them sub-servers) select their own path to reach you. So depending on which ISP and backbone you are connecting through, you will only be disconnected from particular zones. I'm willing to bet money that if you organize all the DC's by zone, the people who DC are in relative proximity or atleast connect through the same backbone or DNS ips are on the same server.

Also I'm not disclosing any header information, because frankly is a bunch of encoded binary. If you can understand it, you can also find it. I don't see a point of me posting it.

 

[Uzor]

Ethernet address? Never heard of it, unless you're talking about MACs. Each device does have a unique hostname and DNS suffix, but that ultimately maps back to an IP.


What he said. BGP is smarter than your average human, and a hell of a lot less lazy. If a link goes down, or is changed, routes are re-established in minutes and within a day the new route is propagated to peers.

YEs, MAC address was the one I was referring to <.<; 4 years and my memory isn't all that clear about all those technical names.

As for when it was, I don't remember. But that too happened right after an update. It always happened right after an update. It'd give us a new route through a faulty backbone. One time it was russia, one tiem it was in Japan. I don't remember where it was the third time, though.

Basically, the backbone were faulty before then, and the timing just happened to be bad luck and it'd make us go through a faulty one.

 

[Arctyrael]

I'm probably disproving myself because the internet works due to its redundancy, but if what you say is true, you route through Russia to hit POL servers.

If that was the trunk (America really only has like 6 MAIN trunks), it is possible you could tracert still to POL servers, but the delay is so high that POL would not allow you to connect.

Thus it isn't a problem of routing - but of a physical innadequacy. The resources just were not available to use.


-edit-
In addition, it sounds like they replaced the line instead of repairing, which accounts for the massive time it took to fix.

 

[Uzor]

/sigh I didn't really want to go into the technical details about it. The servers aren't like based on the server name... There is *no* quetz server. Quetz serve is actually composed of many servers. They are divided by areas mostly. Each server doesn't just hold on area, they hold a few. Each of these (let's call them sub-servers) select their own path to reach you. So depending on which ISP and backbone you are connecting through, you will only be disconnected from particular zones. I'm willing to bet money that if you organize all the DC's by zone, the people who DC are in relative proximity or atleast connect through the same backbone or DNS ips are on the same server.

Also I'm not disclosing any header information, because frankly is a bunch of encoded binary. If you can understand it, you can also find it. I don't see a point of me posting it.

Yeah, that's what I thought. I highly doubt any MMO company is stupid enough to only have 1 computer take care of everything happening on a single server.

 

[Artem]

No Uzor, I wasn't affected. I have family in Russia, they didn't have internet for 3 weeks though.

Also you all are assuming that DNS will refresh itself. Except you are assuming it's a software problem If a vital hard line went down, DNS will keep refreshing but it won't help anything. I think everyone needs to take some time off and be patient. *If* it is a hardline it can take a while. In case of russia the line went out in the middle of Siberia, fixing it in 3 weeks was rather fast for those conditions.

I can't even say what I said is 100% true, that's what I have decided after my own research. Believe me or not. The fact will remain it's down for a while, and you guys need to deal with not being able to go to particular zone. I mean you know you can get out of the zone? Only incoming headers are messed up. Outgoing are fine. you can run through the zone DC'd and when you reconnect you'll be in a new spot.

 

[Myrrh]

Son of a bitch. Now i'm having zone issues where i R0.

 

[Arctyrael]

Also I'm not disclosing any header information, because frankly is a bunch of encoded binary. If you can understand it, you can also find it. I don't see a point of me posting it.

I lawled at that

 

[Everlasting]

Serious question, for the 3rd time.

If this is truly a backbone type issue then why is it only happening to FFXI and nothing else?

 

[Zyndara]

That is the most retarded statement ever. They are working on it for sure. Backbone problems don't get fixed in a day. When the backbone in russia collapsed, it was out for like 3 weeks lol.

No, it isn't at all, though your response is the most pretentious, self-important statement ever. I didn't say they'd get fixed in a day. I don't care how long it takes them to fix it, I care that it gets fixed. What I have a problem with is SE saying "Sorry, not our problem," and not giving us any real indication that this is being looked into seriously or any relative timetable for a solution. Them telling is to try to log in again later isn't a strong indication that they are working hard on fixing it.

 

[Artem]

Serious question, for the 3rd time.

If this is truly a backbone type issue then why is it only happening to FFXI and nothing else?

dunno, SE chooses a crappy way to get some where? And how do we know it's not happening to anything else? Do you know of anything else that connects the same path as a particular zone? I personally don't. Also a packet lost somewhere isn't as bad, as when it is lost in a videogame. If packets are lost in a video game you are gonna be disconnected from the game, in case it's a hacking attempt or anything else of the sort. In other types of software you are just losing a packet that will get resend, and probably will go through a new path to get to you.

 

[Uzor]

Serious question, for the 3rd time.

If this is truly a backbone type issue then why is it only happening to FFXI and nothing else?

Serious answer for the third time, because everything on your computer don't use the same backbone. If you chat with someone on AIM, the backets might not go through the faulty backbone.

And as Artem said, it depends. As I said before, SE have a shitty way of delivering packets. They send every packet twice as a way make sure it reaches its destination. Most companies use more reliable methods.

 

[Ubuntu]

I think SE should provide us free VPN access for the interim lol

The problem might not necessarily be in the application layer. If it's in the 3 or 4 layer like the traffic shaping problem vonage's been having with various ISPs... than... um... idk lol

Anyhow, have anyone try using a Socks5 public proxy from Japan and connect from there? but yeah... would be nice to have VPN access to japan