Too Many Files Open

[faragher]

Describe the Bug

After ~28 days uptime, Reticulum crashes due to too many files being open, breaking ratchets and logging, and eventually crashing the entire transport instance.

To Reproduce

I don't think it wise to wait a month to see if I can duplicate, this is a fundamental issue.

Expected Behavior

File handles closed after use.

Logs & Screenshots

Typical:

Oct 18 05:52:07 reticulum.betweentheborders.com rnsd[553]:   File "/home/xxx/.local/lib/python3.11/site-packages/RNS/Identity.py", line 288, in persist_job
Oct 18 05:52:07 reticulum.betweentheborders.com rnsd[553]: OSError: [Errno 24] Too many open files: '/home/xxx/.reticulum/storage/ratchets/xxx>

System Information

• Debian
• Python 3.11.2
• RNSD 0.7.7

Additional context

Working theory is that there's an error inside a try/except cycle which leaves a handle dangling.

Suggestion: Either migrate to with open('file.ext','r') as f: format or move file.close() out of try: and into finally:

[faragher]

Okay, I found the issue. Identity.py line 332 opens a file which is never closed. I'm halfway through changing Identity.py over into with open so I'm going to finish that up and fix the problem. If that's unacceptable, then it's a quick fix closing the file that was opened on 332.

[liamcottle]

and eventually crashing the entire transport instance.

Heh, I've noticed now and then when I get up in the morning, MeshChat is no longer running and had closed itself overnight. I hadn't added any logging for this yet, but this could be related. Seemed to only happen once a month or so... :D

Looks like your error log is related to linux, whereas I'm running on Windows. So no idea until I investigate further, but thought I'd throw in some extra info...

[markqvist]

Yep, that was a very classical oversight of forgetting to close the file handle again, and even in such an often-used function. Oof! Thanks for the report and fixing it. I'll release an update for this shortly! I'm betting it's the same cause for your crash @liamcottle.

[faragher]

@liamcottle I would assume (a dangerous word) Windows has a similar file handle limit, which may or may not be even the same magnitude as Linux, but I'm willing to bet it would be a problem eventually. Given the similar timing, it wouldn't surprise me, but going back a few versions, sometimes things happened without any error logs I've noticed.

That being said, I can now target these crashes pretty well: I noticed it was down when the Amber Pages said everything was "two days ago." Gave me a ~24 hour period to look at in the logs.

So while this was a problem that has been resolved, I think I have an edge on future troubleshooting, as I think that system is on debug logging. I'll check when I switch over to the new build.