Fix write_to_file function

[gabrielcocenza]

Probably because of the usage of low level functions, as for example os.open and os.fdopen, it's harder to deal with buffering which can result into files with inconsistent content, which breaks the configuration and the hardware-exporter service.

After adding using high level functions and make the code simpler, the bug seems to be fixed

Closes: #305

[jneo8]

Thanks, overall LGTM. Need more information on os.chmod part.

[samuelallan72]

Probably, because of a racing condition,

I feel like if we have race conditions where code is trying to write to the same file at the same time, then we should fix the source, not just add file locking. It may require a refactor to only write out the config from one place. It would be nice to do some more debugging and find the source of the race condition here.

[gabrielcocenza]

I feel like if we have race conditions where code is trying to write to the same file at the same time, then we should fix the source, not just add file locking. It may require a refactor to only write out the config from one place. It would be nice to do some more debugging and find the source of the race condition here.

I tested again and removing the lock file, but keeping the implementation seems to solve the issue. I didn't see in the logs more than one hook triggering the write_to_file function.

I'm not sure about the root cause. My guess is that os.open is a low-level system call. Maybe when you wrap the file descriptor in os.fdopen(), you're attempting to treat a low-level file descriptor as a higher-level file object, but subtle issues can occur if the file descriptor isn't handled perfectly (e.g., file not properly synced, buffers not flushed).

I believe that keeping things simple and using more high level python functions are enough to solve the issue

[samuelallan72]

Thanks for experimenting! It's nice that we can go with something even simpler than the previous implementation and solve the bug at the same time. 👍

[Deezzir]

Legit