Martynas Pumputis
Martynas Pumputis
> Any idea? Without seeing the full verifier log it's difficult to answer. Anyway, my idea was to get rid of `snat_v4_process()` completely instead of making it call `snat_v4_nat() /...
> for 1) is that really makes sense to merge code that did not pass CI? Why would the CI not pass? The suggestion is to create a new tailcall,...
> Have you considered that opportunity? I think we are discussing about the same thing just in different terms. I'm also proposing the split. Just how to get to it....
Converting to draft until #20425 has been resolved.
@sahid It seems to be a legit failure - https://github.com/cilium/cilium/runs/7666956048?check_suite_focus=true#step:8:523.
> One comment is that this would be cool to add a commit description specifying why this change was required. Good point, added.
I don't see how to investigate it without sysdump and traces of each hop.
I still need to know what packet and hops.