Open jsalatiel opened 9 months ago
I have added the support bundle here: https://fastupload.io/bSD9eHRH2c8f0wU/file
@jsalatiel can you check sysctl net.ipv4.ip_forward
on the Nodes? I suspect Openshift doesn't enable it by default. If it's 0
, you may enable it by sysctl -w net.ipv4.ip_forward=1
. If this is the cause, I'm thinking if we should do it by default as it seems relying K8s components to do it seems not working in some cases.
For antctl trace-packet
, it may be a bug, I created https://github.com/antrea-io/antrea/issues/5831 to track it.
Hi @tnqn , it worked , tks! In all my previous tests I was doing a single node installation In that mode the installation would finish and I could SSH to the single node, but I would not get connectivity from the pods as I mentioned in this ticket.
After you mentioned the net.ipv4.ip_forward, i tried using a 3 node cluster. The installation never finishes ( aborts as stalled ). So I destroyed the cluster and created a new one, and I noticed that all the nodes also had net.ipv4.ip_forward=0, I manually set those to net.ipv4.ip_forward=1 in the middle of the installation and the installation finished successfully.
So it would be really nice if antrea could do that net.ipv4.ip_forward=1 by itself mainly because of the readonly nature of redhat core OS.
The remaining problem is that for openshift 4.14.x antrea is not certified thus the third-party collaborative support between Redhat and Vmware wont apply if I use antrea on 4.14. I have opened #99 for that although I have no idea how that certification process works.
Describe the bug
Since I could not find any documentation about how to install antrea on openshift using their new install method (openshift assisted installer) I used Calico's documentation (making the required adjustments) to install antrea as the primary CNI. That basically means configure everything on redhat console panel, including all manifests from the deploy folder and before effectively click "install" issue the following POST.
The installation finishes successful and I can see all pods in running state.
Antrea also appears to be the primary CNI:
The problem is that all pods (not on hostNetwork) have no connectivity to outside the cluster. Pods can connect to themselves, nothing else.
Reproduction steps
Expected behavior
Network should be fine
Additional context
Trace packets fail: