
Post by
zezo | 2017-11-07 | 08:33:05
4 races with 16-day routing are just above my current hardware capacity, so when everyone decided to check the new forecast the system melted down.
It happens like this with servers - a kind of chain reaction. The moment you cross the limit things stars slowing down, people start hitting refresh and pretty soon everything grinds to a halt.
I had similar issues 2 years ago with the 4 Transat boat classes. There was a system in place to bring up google cloud instances on demand that worked pretty well. I have to get it running again - not a big issue, just was not prepared yesterday, and I still haven't observed the actual demand curve - before it was obvious with peaks at 8AM/PM with the weather change. Now the peak seems to be at midnight.