loftar wrote:Masti wrote:First of all, in what intervals are the backups made
Backups are quite irrelevant for the purpose of these kinds of events, because unless I do them constantly (and they can't be done without shutting down the server), they're bound to be older than the data as it was rolled back to anyway.Masti wrote:was the memory overload caused by the game's code or some sort of a DDOS attack or does it perhaps still remain unknown?
In either event, it was caused by the game's code, because a DDoS attack shouldn't cause memory leaks to begin with. Whether, however, there is an exploit that can be used to trigger the bug that causes the server to consume memory rapidly is unknown, however, since I don't know what causes it (or I would have fixed it). I have found no reason to suspect that it is an exploit, however.
The difference in this case, however, is that so much memory was consumed that the kernel's OOM killer went in and reaped the server process before I had a chance to take care of it properly. I am considering turning off memory overcommitting.Masti wrote:Secondly, if the server just crashed, how did some data stay intact while the other parts were, seemingly, copied in from a backup afterwards?
It's simply the last saved version of the data that is used. The data that has been "lost" was simply such data that had not been saved from the live state of the server at the point where the kernel killed it.
instead of buttfucking the server u should give me a pink cape