John Howard
John Howard
I think right now we still hold upstream connections when downstream disconnects. At the very least, we do when we are trying to connect to the upstream; at some point...
We have benchmarks but no tracking. Requirements: * Presubmit should have a way to run against HEAD and compare results. * If this is slow/expensive, we can make it on-demand...
``` $ cargo bench connections/hbone -- --profile-time 10 Finished bench [optimized] target(s) in 0.18s Running benches/throughput.rs (out/rust/release/deps/throughput-9d94640613568b5f) thread 'ztunnel-proxy-2' panicked at 'internal error: entered unreachable code: SecretManager worker died: requests...
Currently, we report these counts only on connection close. We should report them periodically as well, so we get visibility into long running requests
When a user application accepts or initiates a connection, it should be able to somehow access metadata from ztunnel about the connection. At minimum, this would be the identity of...
Istio and Envoy both integrate with OSS fuzz. Figure out how we can, as well. This is likely a bit different since we use rust, but I think there are...