Conversation
Fix various things in the 06-lb test case: --backend does not exist anymore, we need to switch to --backends instead. Same for `cilium service inspect` that should have been `cilium service get` command. Comparing the deletion result from `cilium service delete` has changed to display the id as well. Thus, adapt results to compare against. The `IPV4=1; ./06-lb.sh` subtests seems to have some logical bugs, f.e. we add a legit service with id 10 and later on still expect the `cilium service list` to be empty, but right in the subsequent check we test whether the service with id 10 was the only one present?! The later added service with id 20 has a conflicting frontend vip, so that one cannot succeed. The `cilium endpoint ct dump` doesn't exist anymore, so just comment it out. Add a `cilium service list` after setting up the v4/v6 lb, so we can inspect the mappings. Fixes: c7d4002 ("Switch to spf13/cobra and spf13/viper to implement CLI & config") Signed-off-by: Daniel Borkmann <[email protected]>
| } | ||
|
|
||
| sudo cilium endpoint ct dump $SERVER1_ID | ||
| #sudo cilium endpoint ct dump $SERVER1_ID |
There was a problem hiding this comment.
This is now cilium bpf ct list $SERVER1_ID
There was a problem hiding this comment.
All commands which directly access bpf maps have been grouped in the bpf command to make it clear what uses the remote API and what uses direct access APIs
|
Besides the |
| sudo cilium service update --frontend [::]:80 --backends [::1]:90,[::2]:91 --id 1 --rev 2> /dev/null || { | ||
| abort "Service should have been added" | ||
| } | ||
|
|
There was a problem hiding this comment.
You do not need to change the backend to backends since I fixed it in my lb_wrr review. I also took care of the remaining non working bits in this script. I am fine either ways. If you want to commit this one I will simply ignore this file. I will still need to call it backend though.
This should give us some additional information to debug flakes, e.g. for cilium#361 Signed-off-by: Tobias Klauser <[email protected]>
Add error to the debug output for ipcache check failures. This should help solve cilium#361. Signed-off-by: Jarno Rajahalme <[email protected]>
Add `--skip-ip-cache-check` flag with the default set to true. This is meant to be a temporary flag while we investigate what's causing the flake. Ref: cilium#361 Signed-off-by: Michi Mutsuzaki <[email protected]>
Follow-up for cilium#503 to address cilium/cilium-cli#503 (comment) Also add a comment so we don't forget to re-enable the check again once issue cilium#361 is resolved. Signed-off-by: Tobias Klauser <[email protected]>
* chore: add Cilium graduation blog post * update summary Signed-off-by: Bill Mulligan <[email protected]> --------- Signed-off-by: Bill Mulligan <[email protected]> Co-authored-by: Bill Mulligan <[email protected]>
Fix various things in the 06-lb test case: --backend does not exist
anymore, we need to switch to --backends instead. Same for
cilium service inspectthat should have beencilium service getcommand.Comparing the deletion result from
cilium service deletehas changedto display the id as well. Thus, adapt results to compare against.
The
IPV4=1; ./06-lb.shsubtests seems to have some logical bugs,f.e. we add a legit service with id 10 and later on still expect the
cilium service listto be empty, but right in the subsequent checkwe test whether the service with id 10 was the only one present?!
The later added service with id 20 has a conflicting frontend vip,
so that one cannot succeed. The
cilium endpoint ct dumpdoesn'texist anymore, so just comment it out. Add a
cilium service listafter setting up the v4/v6 lb, so we can inspect the mappings.
Fixes: c7d4002 ("Switch to spf13/cobra and spf13/viper to implement CLI & config")

Signed-off-by: Daniel Borkmann [email protected]
cilium bpf ct list $SERVER1_IDbpf ct listchange, this looks good. I will rebase my other PR on top of this which adds the ability to resolve a hostname to multiple backends if the DNS replies with multiple IP records.