mount: Software caused connection abort

[powellnorma]

I think maybe on slow drive it takes > fs.fuse.max_request_timeout, which results in "Software caused connection abort" which triggers subsequent "Transport endpoint is not connected" stale mounts.

$ borg -V
borg 1.4.4

Linux 6.17.7

[ThomasWaldmann]

How can I reproduce that?

Did you have that with a remote repository?

[powellnorma]

No, just an external HDD that is a bit slow (I believe). Even just borg info takes ~13s (I am not sure if that is more than usual).

Encrypted: No
                       Original size      Compressed size    Deduplicated size
All archives:               62.78 TB             31.58 TB            355.98 GB

                       Unique chunks         Total chunks
Chunk index:                 4339564           1077763052

It uses zstd for compression.

How can I reproduce that?

Maybe cgroups could be used to artificially limit the bandwidth, or via systemd-run:

systemd-run \
    -p IOReadBandwidthMax=/dev/nvme0n1 20M \
    -p IOWriteBandwidthMax=/dev/nvme0n1 10M \
    my_program

Then just try to mount a big borg repo.. and hope it reproduces?

If you want me to give you outputs of any commands / test scripts, I can provide them, just lmk.

Btw, it sometimes works after retrying a couple of times.. --debug --foreground shows:

fuse: _process_archive completed in 53.5 s for [..]

[ThomasWaldmann]

Try the mount again, but run it in foreground (-f), so that you see if it crashes.