TNFS over TCP Bug being investigated. Anyone want to help? :)

[tschak909]

We are investigating a bug with the latest release of the #FujiNet v1.3 firmware, which causes timed out TCP connections to TNFS to not reconnect properly.

 

This error appears when the FujiNet is left on for an extended period of time.

 

If anyone wants to help, code is here: https://github.com/FujiNetWIFI/tnfsd

 

[_The Doctor__]

didn't you say you 'purposely' limited the time for a tnfs connection in previously talk?

If I left a connection to the tnfs server active while searching the directories or mounted a disk from your server and tried the next evening or day it wouldn't work until I did a new trip to your tnfs server and perused it fresh and remounted the image to the slot. This was the behavior before the update.

Edited May 3 by _The Doctor__

[tschak909]

Just now, _The Doctor__ said:

didn't you say you 'purposely' limited the time for a tnfs connection in previously talk?

If I left a connection to the tnfs server active while searching the directories or mounted a disk from your server and tried the next evening or day it wouldn't work until I did a new trip to your tnfs server and remounted the image to the slot. This was the behavior before the update.

UDP connections had a timeout, but it would attempt to recover if the socket would time out.

 

It looks like we need to improve the recovery for TCP connections, as well. (aka keep-alive)

 

-Thom

[cathrynm]

14 hours ago, _The Doctor__ said:

If I left a connection to the tnfs server active while searching the directories or mounted a disk from your server and tried the next evening or day it wouldn't work until I did a new trip to your tnfs server and perused it fresh and remounted the image to the slot. This was the behavior before the update.

I've noticed the same thing too before and after the latest update.This kind of bug can be a pain to diagnose because the repro steps require letting sit all night. For me, just now I saw it.  I had drives on my personal TNFS connected (on APple 2), and then the next morning I can't connect, I can't even do a directory of my own tnfs server, but I can do a directory from apps.irata.online, the other ones. I reset the fujinet and it magically heals itself, I can connect again.

[tschak909]

2 minutes ago, cathrynm said:

I've noticed the same thing too before and after the latest update.This kind of bug can be a pain to diagnose because the repro steps require letting sit all night. For me, just now I saw it.  I had drives on my personal TNFS connected (on APple 2), and then the next morning I can't connect, I can't even do a directory of my own tnfs server, but I can do a directory from apps.irata.online, the other ones. I reset the fujinet and it magically heals itself, I can connect again.

Yup, indeed. This is why I'm letting everybody I can know that this does happen, that we're trying to fix it, and anybody who wants to jump in and help, would be most welcome.

 

-Thom