eric846
eric846
Can we afford to add 5 or 15 counters to help troubleshoot these specific HTTP outcomes? These would be **in addition** to today's catch-all `http_4xx` and `http_5xx` counters. Usually if...
When an Envoy server listens with a custom transport socket, another Envoy downstream with the same transport socket can be the only tool capable of making valid requests to it....
Currently if something goes wrong (e.g. never converging), we lose the whole history of the adaptive session and get only an `absl::Status`. We should return a result proto with results...
Many adaptive load protos specify constraints on field values (`[(validate.rules)...]`), but we never actually call `Envoy::MessageUtil::validate()` to process them. This should be done within `AdaptiveLoadSessionSpecProtoHelperImpl::CheckSessionSpec()`. Note: ``` /** * Validate...
Special percentiles are requested here: https://github.com/envoyproxy/nighthawk/blob/main/source/common/statistic_impl.cc It works fine at the lower percentiles, but 0.990625 seems to always appear instead of 0.99. These are the percentiles from a couple of...
# Investigate strongly typed `ConfigValidator` Can the `ConfigValidator` [0] interface be rewritten somehow so that it knows the plugin-specific proto type, rather than having to try unpacking a Message? [0]...
Currently some integration tests have relaxed requirements about things like byte counts (e.g. >=500). This means when headers produced by Nighthawk or the test server change unexpectedly, the tests continue...
What if Nighthawk had a mode where it tried to send only a handful of requests to the target and dumped the full output? This would also be an opportunity...
- Start an Envoy load balancer listening on 0.0.0.0:12345. - Try to run a load test to http://localhost:12345. You may get connection failures on some machines. Sending the load to...