Well, I think this is resolved and it turns out to be self-inflicted. Yesterday I did a clean Stretch installation in a Qemu VM and did the minimum to get it up so I could do an sshfs mount from it and it too would disconnect the mount after five minutes or so. On a whim I copied it to my laptop running Bullseye this morning and the sshfs connection stayed up for several hours. I then copied the Buster VM over and it too stayed up for well over an hour.
My attention turned to differences in the SSH configuration of each host computer. Several years ago when it appeared I may need to use a Cellular 4G router for Internet access I did not want to lose my ability to SSH into the home LAN remotely. The 4G router provided a less than reliable connect and on the cell network it was placed behind Carrier NAT which meant the IP address it was assigned from the cell network was not reachable from the greater Internet. I set up an AWS host at the time and created an SSH tunnel into it from here that I could attach to with the laptop from elsewhere. The flaky connection caused me to set the sshd_config options of ClientAliveInterval to a value of 300 (five minutes) and ClientAliveCountMax to 1. The idea was to have SSH timeout quickly when the router's 4G connection was reset and I had a script that would attempt a reconnect upon this closure. Commenting these options and letting them be at their defaults of 0 and 3 respectively have apparently resolved my issue. Some months later the WISP built out a new system in this area, which they at first said they weren't going to do that had prompted the 4G experiment. Once they did that I no longer needed the AWS host and SSH tunnel. As I had not seen any issue with connections from over the LAN or Internet with these options set as noted above, I forgot about them, until now. Funny how that works... My guess is that the way Qemu sets up the network bridge that my host could not send the keep-alive message to the guest and simply dumped the connection as it was configured to do, assuming the guest had gone away. In the guest, systemd seeing the closed connection dutifully unmounted the mounts. No big bad bugs to report after all. - Nate -- "The optimist proclaims that we live in the best of all possible worlds. The pessimist fears this is true." Web: https://www.n0nb.us Projects: https://github.com/N0NB GPG fingerprint: 82D6 4F6B 0E67 CD41 F689 BBA6 FB2C 5130 D55A 8819
signature.asc
Description: PGP signature
