503 Service Unavailable from proxy with large number of connections(2.13.4)

[rickwang7712]

There is a forum post which reported about this problem: 503 Service Unavailable from proxy with large number of connections

For my case, I was running performance test with Locust, after the locust users exceeded about 128 users,
the 503 problem will came out with lots of log like this:

[ 265.902625s] INFO ThreadId(01) outbound:proxy{addr=10.244.2.64:7712}:http:rescue{client.addr=10.244.2.65:49176}: linkerd_app_core::errors::respond: HTTP/1.1 request failed error=logical service 10.244.2.64:7712: service unavailable error.sources=[service unavailable]

After checking what the post said above, I tried 2.12.1, but it will hit 500 error(protocol detection timed out after 10s).
To solve error "protocol detection timed out after 10s", I set the opaque port for my service which is described [here](protocol detection timed out after 10s). And then I met the 404 problem mentioned here.
According to the discussion in issue#9888, I upgraded from 2.12.1 to 2.13.4, and then I met 503 error again....

Is there any chance I can solve 503 error?
I used netstat to check my service's connection number, it's about 300~500, I think it's not a large number.

Here is my testing env:

PC A: running minikube with deploying traefik, and 2 svc, one svc handles forwardAuth http request for checking session in redis, another one is a simple http backend which always returns 200.

PC B: runs locust and sends requests to PC A.
ex. locust -f locustfile.py -r 100 -u 1024 --headless -H http://pca-svc:8000 -t 180s --only-summary