-
Notifications
You must be signed in to change notification settings - Fork 440
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix: 1712 - Validate Instrument meta data (name, unit, description) #1713
Conversation
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## main #1713 +/- ##
==========================================
- Coverage 86.28% 85.79% -0.49%
==========================================
Files 169 171 +2
Lines 5160 5212 +52
==========================================
+ Hits 4452 4471 +19
- Misses 708 741 +33
|
OTEL_INTERNAL_LOG_ERROR("Meter::CreateUInt64Counter - failed. Invalid parameters." | ||
<< name << " " << description << " " << unit); | ||
return nostd::unique_ptr<metrics::Counter<uint64_t>>( | ||
new metrics::NoopCounter<uint64_t>(name, description, unit)); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The calling instrumented application will not notice the error if returning a noop counter.
Not crashing now with a noop is not helping.
Instead, bugs will be filed later against opentelemetry-cpp about instrument that produce no data.
How about returning a nullptr, forcing the calling application to do something about it ?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, good point. I too thought about it, but also felt that making critical application crash for some issue in telemetry configuration is not a good idea. We can also be more explicit in error message stating that measurements won't be recorded.
This is also inline with the general error handling guidelines -
Whenever API call returns values that is expected to be non-null value - in case of error in processing logic - SDK MUST return a "no-op" or any other "default" object that was (ideally) pre-allocated and readily available. This way API call sites will not crash on attempts to access methods and properties of a null objects.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for the details. Ok with noop then.
@@ -90,6 +90,14 @@ | |||
# define OPENTELEMETRY_DEPRECATED_MESSAGE(msg) | |||
#endif | |||
|
|||
// Regex support | |||
#if (__GNUC__ == 4 && (__GNUC_MINOR__ == 8 || __GNUC_MINOR__ == 9)) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nit: GNUC_MINOR >= 8?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The regex support is partially implemented for gcc 4.8 and gcc 4.9 only. So the existing condition looks good to me.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
But yeah, even GNUC_MINOR >= 8 should work fine. But let it be explicit. This was already existing in predicate.h, I just moved it here at common place :)
# if HAVE_WORKING_REGEX | ||
: name_reg_key_{kInstrumentNamePattern}, unit_reg_key_ | ||
{ | ||
kInstrumentUnitPattern |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Was this line formatted by clang-format? The format looks a little weird.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yeah clang-format behaved weirdly here. I have disable the clang-format for this block of code. Looks better now.
Thanks for noticing.
kInstrumentUnitPattern | ||
} | ||
# endif | ||
{} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Should this be put under #else
block?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
no it's fine actually. The part of code was somewhat confusing with clang-format. It should be more clear now.
{ | ||
return false; | ||
} | ||
// first char should be alphanumeric |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
First char should be alpha, not alphanumeric?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yeah good point. fixed the comment.
@@ -52,6 +61,14 @@ nostd::unique_ptr<metrics::Counter<double>> Meter::CreateDoubleCounter( | |||
nostd::string_view description, | |||
nostd::string_view unit) noexcept | |||
{ | |||
if (!ValidateInstrument(name, description, unit)) | |||
{ | |||
OTEL_INTERNAL_LOG_ERROR("Meter::CreateDoubleCounter - failed. Invalid parameters." |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nit: probably use macro __func__
to avoid hard code function name in string liternal.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yes, probably we can modify OTEL_INTERNAL_*
macros to automatically add class and method name to the logs. Will create a separate issue for this.
} | ||
} | ||
|
||
TEST(InstrumentMetadataValidator, TestUnit) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Missing unit test on description?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We don't have any validation checks for the description, so no tests added for it.
…ad-local-stack * commit '9acde87b08b225ce511fa8a20c6cba14f2921518': (36 commits) Prepare v1.7.0 release with Metrics API/SDK GA. (open-telemetry#1721) Fix: 1712 - Validate Instrument meta data (name, unit, description) (open-telemetry#1713) Document libthrift 0.12.0 doesn't work with Jaeger exporter (open-telemetry#1714) Fix:1674, Add Monotonic Property to Sum Aggregation, and unit tests for Up Down Counter (open-telemetry#1675) [Metrics SDK] Move Metrics Exemplar processing behind feature flag (open-telemetry#1710) [Metrics API/SDK] Change Meter API/SDK to return nostd::unique_ptr for Sync Instruments (open-telemetry#1707) [Metrics] Switch to explicit 64 bit integers (open-telemetry#1686) Add metrics e2e test to asan & tsan CI (open-telemetry#1670) Add otlp-grpc example bazel (open-telemetry#1708) [Metrics SDK] Add support for Pull Metric Reader (open-telemetry#1701) Fix debug log of OTLP HTTP exporter and ES log exporter (open-telemetry#1703) [SEMANTIC CONVENTIONS] Upgrade to version 1.14.0 (open-telemetry#1697) Fix a potential precision loss on integer in ReservoirCellIndexFor (open-telemetry#1696) fix Histogram crash (open-telemetry#1685) Fix:1676 Segfault when short export period is used for metrics (open-telemetry#1682) Add timeout support to MeterContext::ForceFlush (open-telemetry#1673) Add CMake OTELCPP_MAINTAINER_MODE (open-telemetry#1650) [DOCS] - Minor updates to OStream Metrics exporter documentation (open-telemetry#1679) Fix:open-telemetry#1575 API Documentation for Metrics SDK and API (open-telemetry#1678) Fixes open-telemetry#249 (open-telemetry#1677) ...
Fixes #1712
Changes
For significant contributions please make sure you have completed the following items:
CHANGELOG.md
updated for non-trivial changes