Hi,
I managed to finish it here is my feedback:
In the Dockerfile, from what I have been told and read, it is advisable to do an npm ci
instead of npm install
so that it takes the depencies from the lock.json
and be certain that no version is automatically mounted. One colleague explained this to me yesterday and I found a similar explanation here:
https://stackoverflow.com/a/53325242
In the "Test the app" section, you should specify in the instructions to search around kubectl port-forward
instead of just kubectl
I think.
From this part I had small problems:
To retrieve the hostname, I did the az aks show
except that the command asks for the rg and the name of your aks, you should perhaps specify it in the command to stay consistent with the rest.
And from the result of this command I made the mistake of taking the "fqdn" field as the dns name instead of the one located in HTTPApplicationRoutingZoneName:
"addonProfiles": {
"httpApplicationRouting": {
"config": {
"HTTPApplicationRoutingZoneName": "3f6f8a67980e453ba5e9.francecentral.aksapp.io"
},
…
Apparently the fqdn is that of the plane control.
To retrieve the HTTPApplicationRoutingZoneName
directly, here is the command line that should be specified:
az aks show --resource-group<resource-group-name> --name <cluster-name> --query addonProfiles.httpApplicationRouting.config.HTTPApplicationRoutingZoneName -o table
For debugging I had to check that I had DNS entries in the resource group generated by Azure, maybe say something about it?
Another point in the ingress deployment, the annotations section is deprecated to be replaced by the ingressClassName
field, I tested on aks it seems not yet taken into account on the other hand but interesting to put as a "Note" in the workshop I think :
metadata:
name: web app
# annotations:
# kubernetes.io/ingress.class: addon-http-application-routing
spec:
ingressClassName: addon-http-application-routing # New version
For the metrics recovery part "Get cluster activity metrics", it would be nice to put a screenshot to show where to take the values
On the last part "Enable the Cluster Autoscaler" we launch commands to increase the node pools but we do nothing afterwards, it's a shame I think, why not restart a more brutal Azure Load Testing?
On this occasion in the last script you put Userpool but unless I am mistaken, I did not see a few lines where you explain that there is a Userpool and a Servicepool to separate. Maybe rename it AppWebpool for that matter?
Good Workshop :)