prometheus / prometheus/client_python
Eagerly convert value to float in *MetricFamily.add_metric()
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 4.4k
- Forks
- 876
- Avg merge
- 8d 4h
- Merged PRs (30d)
- 1
Description
Hi there, an experience report and improvement suggestion.
I have a pretty simple exporter which queries a backend URL for JSON, grabs the JSON, stuffs it into prometheus metric families, which it then yields.
Sometimes it errors with stacktraces only when the request is finishing, with a stack traces that points at finish_request rather than the code that added the bad float:
TypeError: ("float() argument must be a string or a number, not 'NoneType'", Metric(bom_wind_speed, Wind speed (km/h) from the Bureau of Meterology, gauge, , [Sample(name='bom_wind_speed', labels={'location': 'Sydney Airport'}, value=None, timestamp=None, exemplar=None), Sample(name='bom_wind_speed', labels={'location': 'Sydney - Observatory Hill'}, value=20, timestamp=None, exemplar=None)]))
Traceback (most recent call last):
File "/usr/local/lib/python3.5/socketserver.py", line 625, in process_request_thread
self.finish_request(request, client_address)
File "/usr/local/lib/python3.5/socketserver.py", line 354, in finish_request
self.RequestHandlerClass(request, client_address, self)
File "/usr/local/lib/python3.5/socketserver.py", line 681, in __init__
self.handle()
File "/usr/local/lib/python3.5/http/server.py", line 422, in handle
self.handle_one_request()
File "/usr/local/lib/python3.5/http/server.py", line 410, in handle_one_request
method()
File "/usr/local/lib/python3.5/site-packages/prometheus_client/exposition.py", line 152, in do_GET
output = encoder(registry)
File "/usr/local/lib/python3.5/site-packages/prometheus_client/openmetrics/exposition.py", line 56, in generate_latest
floatToGoString(s.value),
File "/usr/local/lib/python3.5/site-packages/prometheus_client/utils.py", line 8, in floatToGoString
d = float(d)
TypeError: ("float() argument must be a string or a number, not 'NoneType'", Metric(bom_wind_speed, Wind speed (km/h) from the Bureau of Meterology, gauge, , [Sample(name='bom_wind_speed', labels={'location': 'Sydney Airport'}, value=None, timestamp=None, exemplar=None), Sample(name='bom_wind_speed', labels={'location': 'Sydney - Observatory Hill'}, value=20, timestamp=None, exemplar=None)]))
I've hit this kind of bug a few times in different exporters (I guess it's to be expected to get type errors in Python sometimes).
How about eagerly converting the value passed to add_metric to a float? Then the stack trace would point at the exact cause.
This might be a breaking change - that'd be a reasonable reason to reject this. But any code doing this will likely fail soon after, as soon as an attempt is made to serialize the metric, so maybe it'd be worth the change for better debuggability? What do you think?
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start at MetricFamily.add_metric(), then compare the serialization path in openmetrics/exposition.py and utils.py, including generate_latest and floatToGoString from the traceback. Done means invalid values fail during metric addition rather than serialization, with the compatibility impact and relevant behavior covered by tests.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- prometheus, python
- Domain
- observability-sre
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100