workload entry configure command allows specifying ingress service name #28446
Conversation
google-cla
bot
added
the
cla: yes
istio-testing
added
the
size/S
stevenctl
added
the
release-notes-none
|
/retest |
1 similar comment
|
/retest |
istioctl/cmd/workload.go
Outdated
| ingress, err := kubeClient.CoreV1().Services(istioNamespace).Get(context.Background(), multicluster.IstioIngressGatewayServiceName, metav1.GetOptions{}) | ||
| if err == nil && ingress.Status.LoadBalancer.Ingress != nil && len(ingress.Status.LoadBalancer.Ingress) > 0 { | ||
| ingressIP = ingress.Status.LoadBalancer.Ingress[0].IP | ||
| ingress, err := kubeClient.CoreV1().Services(istioNamespace).Get(context.Background(), multicluster.IstioEastWestGatewayServiceName, metav1.GetOptions{}) |
There was a problem hiding this comment.
since eastwest gateway is just a convention not a requirement should we make this configurable? We will want to support things like external istiod for example, or just a custom gateway setup
There was a problem hiding this comment.
Good point I'll add discoveryServiceName or something similar
There was a problem hiding this comment.
We already allow specifying the IP directly, but service name might be better UX. We do need to give some kind of output here if the IP isn't available (still pending for example).
There was a problem hiding this comment.
does dns name also work or just direct ip? since that will likely be used for external.
we could also take in a hostname instead of service and resolve to service ip if it's internal possibly, although not sure how challenging that is. lot of options
There was a problem hiding this comment.
Right now this is used to create a /etc/hosts "addendum", then in the proxy config we have discovery address as istiod.istio-system... - I figure the DNS name would have to be used as discovery address.
I think tying the gateway to a k8s service is the common case. if they're doing something special we should cover the DNS name case, the manual IP case
As far as resolving a hostname.. not sure how we'd do that? Are we trying to use the address that we resolve to and use the ClusterIP? I figure looking up by name is more reliable. I can see a <svc>.<namespace> like syntax being good so we're not restricted to the system namespace.
There was a problem hiding this comment.
Is this good for now? I'd like to be able to remove the "get ip" step from the docs.
istio-testing
added
the
needs-rebase
stevenctl
force-pushed
the
x-wle-cfg-ew-gw
branch
from
016eb4b to
d8de44a
Compare
istio-testing
removed
the
needs-rebase
stevenctl
changed the title
stevenctl
force-pushed
the
x-wle-cfg-ew-gw
branch
from
d8de44a to
0e8d230
Compare
stevenctl
force-pushed
the
x-wle-cfg-ew-gw
branch
from
0e8d230 to
b77448b
Compare
Co-authored-by: Shamsher Ansari <shaansar@redhat.com>
|
/retest |
|
In response to a cherrypick label: new pull request created: #28584 |
We're now using east-west gateway (soon to be called expansion gateway #28196) for accessing istiod from outside the cluster. The
istioctl x workload entry configurecommand needs to grab the IP from this instead of the ingressgateway.