Cheaper logcontext debug logs (pseudo_random_string(...))

[MadLittleMods]

Cheaper logcontext debug logs (pseudo_random_string(...))

Follow-up to #18966

During the weekly Backend team meeting, it was mentioned that random_string(...) was taking a significant amount of CPU on matrix.org. This makes sense as it relies on secrets.choice(...), a cryptographically secure function that is inherently computationally expensive. And since #18966, we're calling random_string(...) as part of a bunch of logcontext utilities.

Since we don't need cryptographically secure random strings for our debug logs, this PR is introducing a new pseudo_random_string(...) function that uses random.choice(...) which uses pseudo-random numbers that are "both fast and threadsafe".

Dev notes

• https://docs.python.org/3/library/secrets.html
• https://docs.python.org/3/library/random.html

Pull Request Checklist

• Pull request is based on the develop branch
• Pull request includes a changelog file. The entry should:
  • Be a short description of your change which makes sense to users. "Fixed a bug that prevented receiving messages from other servers." instead of "Moved X method from EventStore to EventWorkerStore.".
  • Use markdown where necessary, mostly for code blocks.
  • End with either a period (.) or an exclamation mark (!).
  • Start with a capital letter.
  • Feel free to credit yourself, by adding a sentence "Contributed by @github_username." or "Contributed by [Your Name]." to the end of the entry.
• Code style is correct (run the linters)

[MadLittleMods]

I also considered naming this random_string_insecure_fast

🤷

[reivilibre]

I think we should call this insecure_random_string. Being pseudo-random doesn't necessarily imply insecure, that's what CSPRNGs are for, after all.

It's maybe not a huge deal, but given the sheer hazard of using the wrong type of random in the wrong place, I much prefer the clear and simple insecure label, because it brings your attention to an important (negative) caveat. In theory that could raise some alarm bells at a critical time during review.

On the other hand, I don't think it's important to say fast — if someone consciously thinks about the speed at PR review time, they can look it up. (But I'm also not against calling it 'fast', to be fair!)

[MadLittleMods]

As another alternative, we could gate the random string creation behind if logcontext_debug_logger.isEnabledFor(logging.DEBUG)

[reivilibre]

I think we should call this insecure_random_string. Being pseudo-random doesn't necessarily imply insecure, that's what CSPRNGs are for, after all.

It's maybe not a huge deal, but given the sheer hazard of using the wrong type of random in the wrong place, I much prefer the clear and simple insecure label, because it brings your attention to an important (negative) caveat. In theory that could raise some alarm bells at a critical time during review.

On the other hand, I don't think it's important to say fast — if someone consciously thinks about the speed at PR review time, they can look it up. (But I'm also not against calling it 'fast', to be fair!)