API Services Portal Bug Issue
Seems isolated to Cypress tests. Currently not able to duplicate locally - might just be a matter of adding more wait time.
Failures on two instances of verify rate limit error when the API calls beyond the limit:
Test Case
ENV
TESTCASE
Failures are on tests where:
- global rate limit (20/hr, local policy) is set first,
- then a consumer limit (2/hr, adjusted down to 1/hr following redis key clearing fix) is set (for Service and Route, respectively, in each test).
EXPECTED
- The first request returns 200
- The second request returns 429 Too Many Requests
ACTUAL
- The first request gets a 429
ANALYSIS
The cause seems to be carried over requests from previous tests being stored in redis, leading to a 429 error.
An attempted fix (which works when running Cypress locally) was to clear relevant redis keys via redis-cli before the test requests. This actually works so well (locally at least) that the test needed to be adjusted to the (expected) limit of 1/hr, because the test makes 1 request (200 expected), then a second with 429 expected. Nevertheless, the issue of first request returning 429 persists in CI e2e testing.
Sometimes only 1 of the 2 tests fails, e.g. #1295
API Services Portal Bug Issue
Seems isolated to Cypress tests. Currently not able to duplicate locally - might just be a matter of adding more wait time.
Failures on two instances of
verify rate limit error when the API calls beyond the limit:and
Test Case
ENV
e2eDEVTESTPRODTESTCASE
Failures are on tests where:
EXPECTED
ACTUAL
ANALYSIS
The cause seems to be carried over requests from previous tests being stored in redis, leading to a 429 error.
An attempted fix (which works when running Cypress locally) was to clear relevant redis keys via redis-cli before the test requests. This actually works so well (locally at least) that the test needed to be adjusted to the (expected) limit of 1/hr, because the test makes 1 request (200 expected), then a second with 429 expected. Nevertheless, the issue of first request returning 429 persists in CI e2e testing.
Sometimes only 1 of the 2 tests fails, e.g. #1295