Did this have something to do with the crash that happened an hour or so ago?
I was playing a RT game, my warlight crashed, and so did another player's, we were both booted but the rest weren't affected...
I'm glad some people are finding it's faster. Personally I'm finding it's a tad slower, but I'm still tuning it. It will get better over time. I've been tuning the old host for years so it will take more than one day to get the new one up to the same level.
The site was down for a while a few hours ago. This was bad timing for me. I watched the site like a hawk for the first 16 hours after moving it and everything was working okay, and then as soon as I go to sleep it goes down. It was just really bad timing. It's also supposed to auto-recover from these types of failures and I still need to figure out why the auto-recovery system didn't work.
Also, I just found out that my warlight.net email is down. If you sent me anything at firstname.lastname@example.org in the last 24 hours and I didn't reply, I didn't get it. I'm working at fixing my e-mail now and will make another post when it's back up.
I'm not having this problem anymore, but for the first 24 hours after the server crashed I kept on getting a message saying that my browser had to be refreshed every time I tried to access the my games screen, or that my username and password had just been changed.
Not sure if it has anything with the new host but for some reason i can't access 3 of my games which is weird because i could access all the rest of them. It just sits there loading the game for about 10 minutes then i refresh and try again.
Everything should be working much better now. I've fixed the most critical issues, and will now start working on the less critical ones.
My e-mail is back up. If you sent something to me at email@example.com over the past couple days and didn't get a reply, please re-send it. Outgoing e-mails were not affected.
Technical explanation: I learned never set the root of a domain to a CNAME record. Doing this breaks things, like the MX records needed to make receiving e-mail work. Amazon sort of leads you down the wrong path here since they tell you not to use a standard "A record" with their load balancer, and they tell you that CNAME records work fine. But it's a trap!
There was another issue that map developers ran into that broke the ability to change maps out of development mode. This was caused by a missing dependency on freeglut-dev. I hadn't initially deployed this to production since I didn't realize it was needed. This is now fixed.
There was another issue causing some users to get a session timeout error. After banging my head against the wall for a while I eventually realized this was caused by the missing www redirect. Normally www.warlight.net redirects to warlight.net. This ensures all players are on the same domain so cookies and other things work fine when players send links to each other.
Initially this redirect wasn't working when I migrated, so I disabled it temporarily and made a note to come back to it. I didn't think it would cause any issues, but I was wrong. Players who typed in www.warlight.net into their address bar then loaded flash would have flash running at www.warlight.net but trying to connect to warlight.net. This caused the browser to not pass along the authentication token since it's a cross-domain request which caused their login to be rejected.
I've now put the www redirect back in place so this should be fixed. In the next update I'm also going to make flash always talk to the same domain it's running on so cross-domain requests should never happen in these same circumstance.
Yesterday most of the site slowdowns during the first half of the day were caused by the database machine running out of CPU. It was sitting at 100% utilization which was causing a lot of slowdowns and some time-outs (could-not-connect messages). Normally the WarLight database has always been I/O constrained, so I was surpassed to see it demand so much CPU. It has been upgraded to a faster machine and has been running perfectly ever since.
I've also set up an alarm system on my monitoring application that will make it wake me up at night if the server goes down. The first night the server had a problem while I was sleeping which prolonged the issue. This shouldn't happen again.