Thanks Matt for the infos.
In fact I found the root cause of this problem which was not from the probe itself but from the dns server.
It comes from a conflict between coreDNS and dnsmasq... because I'm trying to remove the virsh layer of the Aiab multinode Site to split the deployment in openstack VMs.
So this is a side-effect of my modifications and a soon as dnsmasq is stopped, the CoreDNS start successfully without modifying the probe.
So I succeed to finish the Genesis Phase and I have a functional Airflow, but as you can figure out I face other problems... to be described in another thread :)

Loïc

Le ven. 17 mai 2019 à 18:02, MCEUEN, MATT <MM9745@att.com> a écrit :

Hey Loic,

 

Good to hear that your multinode testing is progressing well, and great catch on the coredns probe!

 

I think it’s a good idea to err on the side of caution with our coredns probe.  Would you mind submitting a patchset to promenade, making your change here:

https://opendev.org/airship/promenade/src/branch/master/charts/coredns/templates/bin/_probe.sh.tpl#L15

 

That would allow folks to easily weigh in on whether those are the right values?

The patchset will also allow you to easily pull in the change to a AiaB multi-node deployment – you can make a local edit to the AiaB versions.yaml file:

https://opendev.org/airship/in-a-bottle/src/branch/master/deployment_files/global/v1.0demo/software/config/versions.yaml

 

In versions.yaml, find data.charts.coredns, and make these changes:

    Change location to: https://review.opendev.org/airship/promenade

    Change reference to:  the “refs/changes/x/y/z” reference from your promenade patchset

 

Thanks for catching this – it would be awesome if you can resolve it for others.  Let me know if any questions come up or if you need any pointers around the gerrit submit/review process.

 

Thanks,

Matt

 

From: Loic Le Gal <l.legal.astellia@gmail.com>
Sent: Friday, May 17, 2019 5:05 AM
To: airship-discuss@lists.airshipit.org
Subject: [Airship-discuss] CoreDNS problem with liveness/readiness probe

 

Hello Airshipers,

 

After succeeding to deploy a SingleNode config, I'm trying to deploy the multi-nodes-gate site of Airship in a botttle.

The first steps are Ok (Genesis creation and run on node n0) but the Undercloud deployment fails in the creation of coredns pods in kube-system namespace. (Endless Crashloop)

 

I fix manually this problem but want to fix it automatically before launching the whole Airship deployment (i.e. In an Openstack stack) because the initial coredns failure leads to side-effects that needs a new deployment...

 

After investigation, it turns out that the problem comes from a timing issue (may be from the infra I use) and I succeed to fix this manually by modifying the ConfigMap used by CoreDNS's probe and recreate the pods The problem was in the probe.sh file:

 

this command sometimes fails: 

==> dig +trace +time=2 +tries=1 att.com @127.0.0.1

 

Increasing the time and retries fix the problem:

root@n0:/home/ubuntu# kubectl exec coredns-6494d65b66-dk8g9 -n kube-system -- dig +trace +time=2 +tries=1 att.com @127.0.0.1; echo $?

 

; <<>> DiG 9.11.2-P1 <<>> +trace +time=2 +tries=1 att.com @127.0.0.1

;; global options: +cmd

;; connection timed out; no servers could be reached

command terminated with exit code 9

9

root@n0:/home/ubuntu# kubectl exec coredns-6494d65b66-dk8g9 -n kube-system -- dig +trace +time=10 +tries=5 att.com @127.0.0.1; echo $?

 

; <<>> DiG 9.11.2-P1 <<>> +trace +time=10 +tries=5 att.com @127.0.0.1

;; global options: +cmd

;; Received 40 bytes from 127.0.0.1#53(127.0.0.1) in 5000 ms

 

0

 

(Inception inside!) I succeed in patching the Genesis stage to patch the genesis.sh script and finally patch the template file: 

/etc/genesis/armada/assets/charts/coredns/templates/bin/_probe.sh.tpl

 

Unfortunately this file doesn't seems to be the one used to create the Pods' configmap.

 

Can someone tell me the easiest way to patch this readiness probe before after a Git clone and before launching Airship deployment ?

 

BR,

Loïc