doujiang24
doujiang24
@z760087139 1. 最好单独搞个 issue，看起来不是同一个问题 2. > 导出的 heap 会出现 unknown 的情况 这个具体是什么情况呢？文件内容是 unknown 么？
It's not a bug, since response header is already sent to downstream after returing api.Continue in `EncodeHeaders`. you could try return api.StopAndBuffer in `EncodeHeaders`, it will stop sent response header...
SendLocalReply will response to client in DecodeHeader/Data, not data sent to upstream.
通常默认值就可以，具体参数的含义，可以参考 README 文档
Thanks @yanavlasov , we want envoy could quit as soon as possible, and as safe as possible. We are building [AI Gateway for inference scheduling](https://github.com/aigw-project/aigw), the requests are long running...
Thanks @yanavlasov We do not want to wait a new request before closing H/1 connections, it will make us waiting for much longer time. Instead, we hope to close idle...
> What is your current configuration? Envoy start with `envoy --drain-strategy immediate ...`, and run `curl -X POST http://localhost:15000/drain_listeners?graceful` in pre-stop script, thanks @antoniovleonti
> Do you configure an idle timeout? we confiure a short idle timeout(2 min) for now, but we do want to make it much longer, since short idle timeout leads...
> So the issue is you want to increase your idle_timeout without increasing the amount of time spent draining HTTP1 connections from listeners at shutdown? is that right? Yes, that's...
Hi @infor7 , How much total cpu does envoy take? per Go->C call take ~100ns cpu, one `envoyGoFilterLogLevel` call per second shouldn't take too much cpu. Could you probe the...