AMO 🐶⛺🌳

@sabdalla80 , do you think if we force fluentd to reload this will be fixed? This seems like CW logstreams expect to find new names of pods and it is...

@sabdalla80 , thank you for this. I think I understand the problem. The CW plugin requires that the container name be known ahead of time so even though we make...

@sabdalla80 , I still can't produce a valid test with the similar preprocessing block as your config. Can you post the fluentd-config you are using for the namespace and the...

Can you try testing with this image when you have some time? ``` vmwaresaas.jfrog.io/vdp-public/upstream/kube-fluentd-operator:sabdalla80-test-1 ``` EDIT: Image is still being pushed up, give it 2 mins.

@sabdalla80 , I am able to reproduce the preprocessing block now. I see when I scale up from 3 to 6 pods, the preprocessing still stay to 3 blocks...

I think I solved it @sabdalla80 ! By making a function with a channel in the Run() loop (moving it out of RunOnce()) so that it does the check now...

@sabdalla80 I will wait for your test with container before I make PR: ``` vmwaresaas.jfrog.io/vdp-public/upstream/kube-fluentd-operator:sabdalla80-test-2 ```

@sabdalla80 , have you tried running KFO in `--datasource=crd` mode ? Seems to work for me in the crd mode.

@karthiknagraj , currently we don't support this because reloader does not expose any port/endpoints internally. It basically works as a daemon binary. It would be trivial to expose a port...

That should not be that difficult. We can set different livenessProbes for each container I believe.