overly divided network partition about chaos-mesh HOT 3 OPEN

maybe you should take a try with target

https://chaos-mesh.org/docs/simulate-network-chaos-on-kubernetes/#field-description

the network chaos experiment you provided would blackhole each "foo" pod, and also block the connections between "foo" pods.

from chaos-mesh.

I don't know if this matters, but:

1. The labeled worker Pods are from per-Pod StatefulSets and are required to run on specific, separate Nodes, which are tied to specific VMs.
2. The worker StatefulSets are all associated with a headless service.
3. All of the worker Pods have a per-Pod LoadBalancer for most of their TCP ports, exposing those ports externally.
4. All worker Pods have a LoadBalancer for externally exposing two remaining TCP ports.
5. There is one Pod, bound to a separate Node tied to a separate VM, that doesn't have the relevant label. It has its own externally facing LoadBalancer. I haven't checked what happens with that Pod when the 4-way partition is in effect.

from chaos-mesh.

A target selector was needed.

from chaos-mesh.