John Howard

I think you want per workload, the use case I was interested in was putting something that Istio proxy depends on first (bootstrap certs, etc) which is likely per workload....

Yeah I initially tried to pass pipes to it then give go `/dev/fd/3`. I don't really know why it doesn't work TBH, but the reads from the same pipe in...

This really showcases https://github.com/golang/go/issues/61233 well. So much time spent "running" packages with no tests at all!  Spent some time refining the UI

The 5s timeout is already pretty outrageously long (and is the result of retrying! just the hard limit). Did we find it was simply not waiting long enough? it seems...

It seems like this was related to other issues that have been fixed since then. If others are seeing similar please upgrade to 1.24.2+, or open a new issue if...

> > The weird thing is I do not see the same behavior. > > You don't see the same behavior where? :facepalm: I updated the comment

Ah.. I think I just disabled pooling improperly. The benchmark doesn't pool due to using different client network namespaces. I can reproduce full now and see good scaling with number...

Its still pretty interesting that 1->2 yields no improvement, but we do still scale up to some extent. 16c gives 15Gb/s which is consistent with the non-pooling...

|C|Pool|No Pool| |-|-|-| |1|9.27|11.05| |2|10.11|**17.17**| |3|10.60|17.20| |4|11.72|16.24| |6|13.55|14.84| |8|14.34|15.95| |16|16.09|14.71| Actually kind of interesting that we see inverse behavior 

`ZTUNNEL_WORKER_THREADS`, we default to 2. Ztunnel is only connecting to itself in some of the tests, the ones that are meaningful (the ones running in k8s) is cross-node.