Update perf_profiler_bpf_test to use UnitConnector.#1283
Conversation
Signed-off-by: Pete Stevenson <jps@pixielabs.ai>
091704f to
fd698ee
Compare
| std::vector<std::unique_ptr<ContainerRunner>> sub_processes_; | ||
| }; | ||
|
|
||
| template <bool FastTest> |
There was a problem hiding this comment.
Could you explain fast tests vs slow tests? I don't really understand what we're trying to achieve with the distinction.
There was a problem hiding this comment.
The "slow" test accepts all the default profiler parameters, but with the 30 sec. data sampling periodicity, the test inherently runs slower. The "fast" test overrides the defaults, but the test case will complete in less time. At some point in the history, we had (or were going toward) all tests using the "fast" methodology but a reviewer suggested that we keep at least one test with the slow/default setup. This PR does not change the existing setup, but it does provide a nicer way to identify which test is in which category by naming the templated test fixtures accordingly.
Summary: we convert `perf_profiler_bpf_test` to use `UnitConnector` to share code across different tests and binaries. We also refactor the Java test case to separately test the expectation that Java symbolization artifacts are removed after processes stop. The main Java test case then conforms to the pattern established for Go and cpp. We also update some of the helper functions in the test to tidy up the common test cases. Type of change: /kind cleanup Test Plan: existing tests & runs per test. - [x] `bazel test //src/stirling/source_connectors/perf_profiler/... --config bpf --runs_per_test 16` ([all tests passed](https://phab.corp.pixielabs.ai/P372)). --------- Signed-off-by: Pete Stevenson <jps@pixielabs.ai>
Summary: we convert `perf_profiler_bpf_test` to use `UnitConnector` to share code across different tests and binaries. We also refactor the Java test case to separately test the expectation that Java symbolization artifacts are removed after processes stop. The main Java test case then conforms to the pattern established for Go and cpp. We also update some of the helper functions in the test to tidy up the common test cases. Type of change: /kind cleanup Test Plan: existing tests & runs per test. - [x] `bazel test //src/stirling/source_connectors/perf_profiler/... --config bpf --runs_per_test 16` ([all tests passed](https://phab.corp.pixielabs.ai/P372)). --------- Signed-off-by: Pete Stevenson <jps@pixielabs.ai>
Summary: we convert
perf_profiler_bpf_testto useUnitConnectorto share code across different tests and binaries.We also refactor the Java test case to separately test the expectation that Java symbolization artifacts are removed after processes stop. The main Java test case then conforms to the pattern established for Go and cpp.
We also update some of the helper functions in the test to tidy up the common test cases.
Type of change: /kind cleanup
Test Plan: existing tests & runs per test.
bazel test //src/stirling/source_connectors/perf_profiler/... --config bpf --runs_per_test 16(all tests passed).