The day the counters were reset
On the EL server, there are a few millisecond counters. Some are used for the walk requests, to prevent lots of walk requests from the same player in a short interval. For example, only 3 walk requests per second. Otherwise someone could send 1K walk requests, which would slow down the server a lot.
Another MS counter is for the harvesting, to prevent people from harvesting too fast.
One problem with those counters is that they are 32b values, which means that in an interval 2^32/(1000*60*60*24)= ~49 days they will reset (overflow, then start from 0).
We never had the server up for that long, so there was a concern on possible problems that might happen when they overflow.
Because of the modifications of the test server for the update, and because of the fact that no critical bugs were discovered in the current server, we just let it run without restarting it. Yesterday it happened that the server was up for 49 days, so the counters restarted.
As expected, there were a few problems, such as people not being able to walk or harvest. However, when they disconnected and reconnected, those problems were solved.
So at least now we know that this is not a really big problem, which is pretty reassuring.
In other news, we released the updated Windows client, so that people can predownload it, which will avoid a big strain on the server that host our update/download files. The server is still strained a little bit, because almost 200 people download the update now, but that's better compared to 500 people downloading it at once. Besides, this way, at the time of the update people with slow connections won't have to wait and download it, as they already have the client. All they will have to do is install it.
Another MS counter is for the harvesting, to prevent people from harvesting too fast.
One problem with those counters is that they are 32b values, which means that in an interval 2^32/(1000*60*60*24)= ~49 days they will reset (overflow, then start from 0).
We never had the server up for that long, so there was a concern on possible problems that might happen when they overflow.
Because of the modifications of the test server for the update, and because of the fact that no critical bugs were discovered in the current server, we just let it run without restarting it. Yesterday it happened that the server was up for 49 days, so the counters restarted.
As expected, there were a few problems, such as people not being able to walk or harvest. However, when they disconnected and reconnected, those problems were solved.
So at least now we know that this is not a really big problem, which is pretty reassuring.
In other news, we released the updated Windows client, so that people can predownload it, which will avoid a big strain on the server that host our update/download files. The server is still strained a little bit, because almost 200 people download the update now, but that's better compared to 500 people downloading it at once. Besides, this way, at the time of the update people with slow connections won't have to wait and download it, as they already have the client. All they will have to do is install it.
posted by Radu at 2:39 PM
1 Comments:
Thanks radu for bringing up this issue. I was online at that time, and it was really strange. I was scratching my head so hard it almost hurt for any possible explanation; such riddles are very tempting for the technically inclined. After the first three lines I would have been able to stop reading, the puzzle parts just fell into the right places .. :-)
Post a Comment
<< Home