At ITA Software, we had servers that the customer airlines would connect to, and they would sometimes get TCP connection resets for no apparent reason. (This was back in the 'aughts, when companies still ran their own servers.) ITA had bought a super-expensive router with back-up power supplies and lots of administration features.
So, we configured logging. Mysteriously, resets increased the next day. So we added more logging. On the third day, the whole system collapsed. Turns out the resets were coming from the super-expensive router itself because it was getting overloaded. More logging meant more load. Ooops!
After that, ops needed CEO approval to turn on any logging. Good times!
So, we configured logging. Mysteriously, resets increased the next day. So we added more logging. On the third day, the whole system collapsed. Turns out the resets were coming from the super-expensive router itself because it was getting overloaded. More logging meant more load. Ooops!
After that, ops needed CEO approval to turn on any logging. Good times!