Replying to myself with some delay admittedly ;-) The issue had nothing to do with Linkerd, but was basically a class loading issue due to a missing env var in the taskmanager.
In case you are interested in the details, and some clarifications on how Linkerd needs to be setup with Flink, see my detailed answer below: - https://stackoverflow.com/questions/76033732/flink-kubernetes-and-linkerd/76320230#76320230 Thanks! On 2023/01/09 15:07:51 Salva Alcántara wrote: > I've been deploying a job without issues until injecting the linkerd proxy > in order to access some services under a service mesh. In particular, I'm > receiving the following error on the job manager side: > > ``` > java.lang.ClassNotFoundException: a.class.of.Mine > ``` > > On the taskamanager side, it cannot join the cluster based on the following > error: > ``` > Could not find any IPv4 address that is not loopback or link-local. Using > localhost address > ``` > > That is weird and I'm assuming a problem with the setup of linkerd since > the jar actually contains the the reported class (`a.class.of.Mine`). In > order to provide some more context, the job is deployed to k8s via the VVP > (version 2.7.1 for what it's worth). > > Regarding the jobmanager/taskmanager templates, I'm using the following pod > annotations: > > ``` > linkerd.io/inject: enabled > config.linkerd.io/skip-outbound-ports: 8125,8126 # for metrics > config.linkerd.io/proxy-await: enabled > ``` > > PS: Also posted in SO: > - > https://stackoverflow.com/questions/75059165/classnotdeffouncexception-when-deploying-flink-in-k8s-using-linkerd >
