so experience shows 2 minutes works better, but not perfect. hard to determine if this is because connection is actually dying or just braindeadness on their side.
recommendation: every connection should keep track of the longest delta between bytes, when a connection dies after a timeout, can print the longest delta that didn't cause a timeout. if said delta is nowhere near the timeout length, indicative that its actually a problem, if the longest delta recorded is somewhat near the timeout length, perhaps value in increasing timeout.
thoughts?
|