I would like to apologize to anyone who was unable to connect, or had difficulty connecting, to yesterday's game. This was due to a problem with the Knights server list system.
The rest of this post contains some further explanation, for those who are interested in what went wrong exactly.
The Knights server system is split into two components:
- The Knights game server itself, which runs on a machine in my home.
- The server list database, which runs on an external server (solarflare.org.uk, which is actually my personal web site -- I pay an external company to host this). This is responsible for displaying the "server list" page on the Knights website, as well as the server list that appears in-game when you click Connect to Server.
Yesterday, the hosting company reported they were having problems that led to "intermittent access slow down". It seems that this prevented people from accessing the "Connect to Server" screen and therefore they were unable to connect. However there was nothing wrong with the Knights server itself, so once you did manage to connect, you were fine.
It appears that the problems lasted for only 1-2 hours but unfortunately it coincided almost exactly with our weekly Knights game -- that's just bad luck I guess :(
I was thinking about how to prevent this happening in future, I thought of two things:
- I should probably move the server list system from solarflare.org.uk to knightsgame.org.uk. There is no good reason to have it on an external server (it's only there for historical reasons really) and having it there just creates another possible point of failure.
- The Knights client could be made more robust. If looking up the server list fails, the client doesn't deal with it very well; I think it just freezes up for a couple of minutes then quits back to the main menu. Instead, it should timeout after a shorter time, and if getting the server list fails, we could provide a default list or maybe just invite the player to type in a server name manually.
Anyway, I will look at point #1 this weekend because it should be fairly quick to implement. (It just needs me to set up a new database and move some PHP files across.) #2 is a bit harder to do because it requires actual programming, so I will probably just put this on the to-do list for now.
Update: I have now moved the server list system from solarflare.org.uk to knightsgame.org.uk. (It took about 10 minutes.) This should prevent last week's connection problem from happening again. (Hopefully!)