Revert "loader : Log upsert and remove route errors"

[nbusseneau]

This reverts commit 32284cd.

It seems like PR #15339 introduced a regression on Kernel 4.9 with the two following tests failing consistently (not flakes):

• K8sDatapathConfig Encapsulation Check vxlan connectivity with per-endpoint routes
• K8sDatapathConfig Host firewall With VXLAN and endpoint routes

Fixes #15499

[nbusseneau]

Note: if we end up merging this, we should re-open #15282 and ping the author of #15339.

[nbusseneau]

Pinging @nathanjsweet to let him know and avoid confusing him, since he merged the original PR ^^

[nbusseneau]

I'm wondering if the issue might be due to the if/else blocks logic having changed.

Original:

cilium/pkg/datapath/loader/loader.go

Lines 362 to 366 in 72e6238

	if ep.RequireEndpointRoute() {
	upsertEndpointRoute(ep, *ip.IPNet(128))
	} else {
	removeEndpointRoute(ep, *ip.IPNet(128))
	}

Merged in #15339:

cilium/pkg/datapath/loader/loader.go

Lines 372 to 380 in 32284cd

	if ep.RequireEndpointRoute() {
	if err := upsertEndpointRoute(ep, *ip.IPNet(128)); err != nil {
	scopedLog.WithError(err).Warn("Failed to upsert route")
	} else {
	if err := removeEndpointRoute(ep, *ip.IPNet(128)); err != nil {
	scopedLog.WithError(err).Warn("Failed to remove route")
	}
	}
	}

If keeping the same blocks, it should have been instead:

if ep.RequireEndpointRoute() {
	if err := upsertEndpointRoute(ep, *ip.IPNet(128)); err != nil {
		scopedLog.WithError(err).Warn("Failed to upsert route")
	}
} else {
	if err := removeEndpointRoute(ep, *ip.IPNet(128)); err != nil {
		scopedLog.WithError(err).Warn("Failed to remove route")
	}
} 

[aanm]

pinging @h3llix @pchaigno. The code point out by @nbusseneau seems the reason why tests were failing.

[aanm]

it seems now that only are failing:
https://jenkins.cilium.io/job/Cilium-PR-K8s-1.15-kernel-4.9/137 is #15337
https://jenkins.cilium.io/job/Cilium-PR-K8s-1.19-kernel-4.19/5159 is #15455
https://jenkins.cilium.io/job/Cilium-PR-K8s-1.13-net-next/1236/ seems to be a new unrelated flake

The remaining CI is failing because this PR was not rebased after #15487 was merged

merging...

[pchaigno]

Oh, whoa! How did I miss that?

@h3llix Do you want to take care of resending a fixed pull request?

[h3llix]

@pchaigno yes i ll do it . Do i have to rebase it ?

[pchaigno]

Yep, rebase, fix the issue, and resend :-)