Data center travel suspended

[localareanetwork]

We had 24 hour maintenance for the thing to break instantly

[Tyrn]

SE's load testing and the load they experience on launch days never seem to match up, either by accident (they just keep failing to predict correctly) or by design (they know the load on launch will be much higher than normal and don't care). Either way I expect the same thing to happen with the island instances in 6.2, best not to touch it right away.

[Zer0ofTime]

SE must have the same QA testing team as Cyberpunk2077 and Battlefield2042

[TaleraRistain]

Regardless of how much testing they would do, the first thing people are always going to do is dogpile a new feature, and this will always cause congestion to happen, especially when practically everyone and their grandmother dogpile a new feature. Managing that congestion is not very easy to do, especially when the number of requests potentially far exceeds the capacity that the servers can handle. This is basically akin to the community doing a DDoS attack.

It's also entirely possible to do heavy stress testing in a development environment but then have something unexpected manifest when it's released to the live environment. You try to plan for us much as you can and have a plan for rolling back if necessary if something unexpected happens.

[Kaurhz]

It's also entirely possible to do heavy stress testing in a development environment but then have something unexpected manifest when it's released to the live environment. You try to plan for us much as you can and have a plan for rolling back if necessary if something unexpected happens.

Indeed, the thing that people seem to forget is that simulations are often done in accordance with abstract models and these often don't align very well with the outcomes you find when doing it in a real, live environment, so whilst you may be able to simulate the load, you won't necessarily catch the issues that you might otherwise find when doing it in a live environment.

I sympathize with people being unable to log in or being caught on a DC as a visitor when they may otherwise be set for raiding or have plans with friends. But honestly let's face it, there was a fairly good chance this feature was going to crash in some way or another.

To those that were complaining about the Monday release schedule, this is why.

[TaleraRistain]

Indeed, the thing that people seem to forget is that simulations are often done in accordance with abstract models and these often don't align very well with the outcomes you find when doing it in a real, live environment, so whilst you may be able to simulate the load, you won't necessarily catch the issues that you might otherwise find when doing it in a live environment.

Not to mention the system you're working on may be fine, but it may or may not be connected to the great hive of systems that could affect things. Like what happened with the lottery server and the housing server. Sometimes there isn't time or resources to completely mock up exactly the same environment. I know I certainly have times I say "I have tested X, Y, Z, A, B" and then acknowledge "C" could still happen and develop a plan for that.

I definitely sympathize with anyone unable to log in and play or are stuck. But I also sympathize with the developer side because I've been in those shoes having to roll back something that didn't do quite what was expected. That they were able to suspend it so fast says to me that they did have a plan and were watching for excessive activity outside their predictions.

[Jeeqbit]

SE's load testing and the load they experience on launch days never seem to match up, either by accident (they just keep failing to predict correctly) or by design (they know the load on launch will be much higher than normal and don't care). Either way I expect the same thing to happen with the island instances in 6.2, best not to touch it right away.

The second one. They know the load will be much higher than normal, but they know it will get better after the initial congestion. There is nothing SE can do to manage unlimited load, unless they invest in cloud servers, but cloud servers are slower and SE decided against them.

We were warned about the initial congestion and it was obvious it could happen.

[DynnDiablos]

Question - so if 15,000 people are trying to visit Zalera for example (my homeworld)... ME who lives on Zalera has to wait in a queue of 15,000+ to login to my own homeworld? How is this even remotely fair to those of us with no interest in DC travel and we get stuck for hours trying to go into our damn world?

[Ransu]

SE's load testing and the load they experience on launch days never seem to match up, either by accident (they just keep failing to predict correctly) or by design (they know the load on launch will be much higher than normal and don't care). Either way I expect the same thing to happen with the island instances in 6.2, best not to touch it right away.

They test expected loads over normal usage. What's going on today is everyone trying to use the feature all at once...which no amount of preparation could ever really prepare for because that's just simply how hardware and network limitations work and they are not going to spend money on the hardware overhead for a problem that will only last a few days.

[hydralus]

Couldn't they have a system in place for handling an overload of requests? Like "Hey there's already 1000 of you trying to hop datacenters right now, the other 14k of you who are trying will have to wait."