Generally speaking, if one side only has something generic about “connection closed”, the root cause will be reported on the other side (that did the closing). If both sides just see a connection close or reset, you should suspect something in between such as a firewall.
I believe we should increase the default timeout. Slower devices (I’m using one) frequently run into this problem which means other machine start using a lot of CPU+IO (to exchange indexes again?).
I disagree on the basis that it’s probably minority of the devices.
Also I’d rather get the blocking part solved so that we wouldn’t need to make these sort of sacrifices.
boltdb release seems to fix the timeout issue for me, I have it now for over a week on my raspberry pi and the issue seems to be gone even with the default 30s timeout…