- Authors: your names here
- Date: 2000-01-01
- Status: draft|review|accepted|closed
- Discussion
A paragraph concisely describing the problem you need to solve and and your recommended solution. A tl/dr as the kids say. For example:
The Agent should be able to run checks written in Python. This is a proposal to
embed CPython using cgo, so checks would run within the same process as the
Agent itself.
If necessary, add a more detailed sketch of the problem. Add any notes, pictures, text or details that illuminate why this is a problem we need to solve. Keep it as concise as possible.
If necessary, note any constraints or requirements that any solution must fulfill. For example:
1. We have to target Python 2.7.
2. Memory footprint should not increase more than 10%.
3. Every check in `integrations-core` and `integrations-extra` should be
supported.
Enumerate possible solutions to the problem. Note your recommended solution first. For example:
Describe your solution in the bare minimum detail to be understood. Explain why it's better than what we have and better than other options. Address any critical operational issues (failure modes, failover, redudancy, performance, cost). For example:
Embedding CPython is a well known, documented and supported practice which is quite common for C applications. We can leverage the same C api using cgo...
- Strengths
- easily share memory between Go and Python
- Weaknesses
- adapt Go's concurrency model to Python threading execution model might cause issues
- we can't cross compile anymore
- a Python crash would bring down the Agent
- Performance
- ...
- Cost
- we add about 4Mb of memory footprint
- ...
Describe a few other options to solve problem.
Jot a few nots on other things you've considered. At each collection cycle the Agent would fork a Python process running a check...
Note any big questions that don’t yet have an answer that might be relevant.
Link any relevant stats, code, images, whatever that work for you.