tinc - May 2017 - How to improve/resolve the Flushing xxx bytes to node would block?

Hi, Guus

I encounter the below log messages that after tinc run for a while, it seems
box1 begin to buffer traffic to abc, and box1 is behind a firewall to connect
abc from firewall’s inside to outside, and abc is on the public internet.

Normally the major traffic is actually from abc to box(download direction), but
not sure how this happens, and could you suggest something we can avoid this to
happen?

Since this happened, the connection will be closed, and give to my setup, the
connection to switch to another node, but I’m looking for ways to get it
improved.

Like, if the box1 deployed outside the firewall direct to the internet will be a
way out?, and could you explain in detail what the Flushing xxx to bbb would
block mean, and what cause it to happen?

May 24 15:50:47 box1 tincd: Sending PACKET to abc (202.45.97.3 port 443)
May 24 15:50:47 box1 tincd: Flushing 12348 bytes to abc (202.45.97.3 port 443)
would block
May 24 15:50:47 box1 tincd: Sending PACKET to abc (202.45.97.3 port 443)
May 24 15:50:47 box1 tincd: Flushing 12420 bytes to abc (202.45.97.3 port 443)
would block
May 24 15:50:47 box1 tincd: Sending PACKET to abc (202.45.97.3 port 443)
May 24 15:50:47 box1 tincd: Flushing 13820 bytes to abc (202.45.97.3 port 443)
would block
May 24 15:50:47 box1 tincd: Sending PACKET to abc (202.45.97.3 port 443)
May 24 15:50:47 box1 tincd: Flushing 15220 bytes to abc (202.45.97.3 port 443)
would block
May 24 15:50:53 box1 tincd: abc (202.45.97.3 port 443) could not flush for 6
seconds (15220 bytes remaining)
May 24 15:50:53 box1 tincd: Closing connection with abc (202.45.97.3 port 443)
May 24 15:50:53 box1 tincd: Could not set up a meta connection to abc


Bright sent from gmail.