Reputation: 417
I have Istio installed on docker-desktop. In general it works fine. I'm attempting to setup an http-based match on a very simple virtual service, but I'm only able to get 404s. Here are the technical details.
My endpoint image is hashi http-echo which uses the net/http
library to create a trivial http server that returns a message you supply. It works just fine and couldn't be more trivial.
Here is my pod and service configuration:
kind: Pod
apiVersion: v1
metadata:
name: a
labels:
app: a
version: v1
spec:
containers:
- name: a
image: hashicorp/http-echo
args:
- "-text='this is service a: v1'"
- "-listen=:6789"
---
kind: Service
apiVersion: v1
metadata:
name: a-service
spec:
selector:
app: a
version: v1
ports:
# Default port used by the image
- port: 6789
targetPort: 6789
name: http-echo
And here is an example of the service working by my curling it from another pod in the same namespace:
/ # curl 10.1.0.29:6789
'this is service a: v1'
And here's the pod running in the docker-desktop cluster:
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
a 2/2 Running 0 45h 10.1.0.29 docker-desktop <none> <none>
And here is the service registering and administrating the pod:
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE SELECTOR
a-service ClusterIP 10.101.113.9 <none> 6789/TCP 45h app=a,version=v1
Here is my istio istio-ingressgateway pod specification via Helm (seems to work fine) which I list as this is the only part of the installation I've changed and the change itself is utterly trivial (just add a single new port block which seems to work fine as listening is indeed occurring):
gateways:
istio-ingressgateway:
name: istio-ingressgateway
labels:
app: istio-ingressgateway
istio: ingressgateway
ports:
- port: 15021
targetPort: 15021
name: status-port
protocol: TCP
- port: 80
targetPort: 8080
name: http2
protocol: TCP
- port: 443
targetPort: 8443
name: https
protocol: TCP
- port: 6789
targetPort: 6789
name: http-echo
protocol: TCP
And here is the kubectl get svc
on the istio-ingressgateway just to show that indeed I have an external-ip and things look normal:
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE SELECTOR
istio-ingressgateway LoadBalancer 10.109.63.15 localhost 15021:30095/TCP,80:32454/TCP,443:31644/TCP,6789:30209/TCP 2d16h app=istio-ingressgateway,istio=ingressgateway
istiod ClusterIP 10.96.155.154 <none> 15010/TCP,15012/TCP,443/TCP,15014/TCP 2d16h app=istiod,istio=pilot
Here's my virtualservice:
apiVersion: networking.istio.io/v1alpha3
kind: VirtualService
metadata:
name: a-service
namespace: default
spec:
hosts:
- 'a-service.default.svc.cluster.local'
gateways:
- gateway
http:
- match:
- port: 6789
route:
- destination:
host: 'a-service.default.svc.cluster.local'
port:
number: 6789
Here's my gateway:
apiVersion: networking.istio.io/v1alpha3
kind: Gateway
metadata:
name: gateway
namespace: default
spec:
selector:
app: istio-ingressgateway
servers:
- port:
number: 6789
name: http-echo
protocol: http
hosts:
- 'a-service.default.svc.cluster.local'
And then finally here's a debug log from the istio-ingressgateway showing that despite all these seemingly correct pod, service, gateway, virtualservice and ingressgateway configs, the ingressgateway is only return 404s:
2021-09-27T15:34:41.001773Z debug envoy connection [C367] closing data_to_write=143 type=2
2021-09-27T15:34:41.001779Z debug envoy connection [C367] setting delayed close timer with timeout 1000 ms
2021-09-27T15:34:41.001786Z debug envoy pool [C7] response complete
2021-09-27T15:34:41.001791Z debug envoy pool [C7] destroying stream: 0 remaining
2021-09-27T15:34:41.001925Z debug envoy connection [C367] write flush complete
2021-09-27T15:34:41.002215Z debug envoy connection [C367] remote early close
2021-09-27T15:34:41.002279Z debug envoy connection [C367] closing socket: 0
2021-09-27T15:34:41.002348Z debug envoy conn_handler [C367] adding to cleanup list
2021-09-27T15:34:41.179213Z debug envoy conn_handler [C368] new connection from 192.168.65.3:62904
2021-09-27T15:34:41.179594Z debug envoy http [C368] new stream
2021-09-27T15:34:41.179690Z debug envoy http [C368][S14851390862777765658] request headers complete (end_stream=true):
':authority', '0:6789'
':path', '/'
':method', 'GET'
'user-agent', 'curl/7.64.1'
'accept', '*/*'
'version', 'TESTING'
2021-09-27T15:34:41.179708Z debug envoy http [C368][S14851390862777765658] request end stream
2021-09-27T15:34:41.179828Z debug envoy router [C368][S14851390862777765658] no cluster match for URL '/'
2021-09-27T15:34:41.179903Z debug envoy http [C368][S14851390862777765658] Sending local reply with details route_not_found
2021-09-27T15:34:41.179949Z debug envoy http [C368][S14851390862777765658] encoding headers via codec (end_stream=true):
':status', '404'
'date', 'Mon, 27 Sep 2021 15:34:41 GMT'
'server', 'istio-envoy'
Here's istioct proxy-status:
istioctl proxy-status ⎈ docker-desktop/istio-system
NAME CDS LDS EDS RDS ISTIOD VERSION
a.default SYNCED SYNCED SYNCED SYNCED istiod-b9c8c9487-clkkt 1.11.3
istio-ingressgateway-5797689568-x47ck.istio-system SYNCED SYNCED SYNCED SYNCED istiod-b9c8c9487-clkkt 1.11.3
And here's istioctl pc cluster $ingressgateway:
SERVICE FQDN PORT SUBSET DIRECTION TYPE DESTINATION RULE
BlackHoleCluster - - - STATIC
a-service.default.svc.cluster.local 6789 - outbound EDS
agent - - - STATIC
istio-ingressgateway.istio-system.svc.cluster.local 80 - outbound EDS
istio-ingressgateway.istio-system.svc.cluster.local 443 - outbound EDS
istio-ingressgateway.istio-system.svc.cluster.local 6789 - outbound EDS
istio-ingressgateway.istio-system.svc.cluster.local 15021 - outbound EDS
istiod.istio-system.svc.cluster.local 443 - outbound EDS
istiod.istio-system.svc.cluster.local 15010 - outbound EDS
istiod.istio-system.svc.cluster.local 15012 - outbound EDS
istiod.istio-system.svc.cluster.local 15014 - outbound EDS
kube-dns.kube-system.svc.cluster.local 53 - outbound EDS
kube-dns.kube-system.svc.cluster.local 9153 - outbound EDS
kubernetes.default.svc.cluster.local 443 - outbound EDS
prometheus_stats - - - STATIC
sds-grpc - - - STATIC
xds-grpc - - - STATIC
zipkin - - - STRICT_DNS
And here's istioctl pc listeners on the same ingress:
ADDRESS PORT MATCH DESTINATION
0.0.0.0 6789 ALL Route: http.6789
0.0.0.0 15021 ALL Inline Route: /healthz/ready*
0.0.0.0 15090 ALL Inline Route: /stats/prometheus*
And finally here's istioctl routes:
NOTE: This output only contains routes loaded via RDS.
NAME DOMAINS MATCH VIRTUAL SERVICE
http.6789 a-service.default.svc.cluster.local /* a-service.default
* /stats/prometheus*
* /healthz/ready*
I've tried numerous different configurations from changing selectors, to making sure port names match to trying different ports. If I change my virtualservice from http to tcp the port match works great. But because my ultimate goal with this is to do more advanced header-based matching I need to be matching on http. Any insight would be greatly appreciated!
Upvotes: 1
Views: 1611
Reputation: 417
It turned out the problem was that I had specified my service in my hosts
directive in both my gateway and virtualservice. Specifying a service as a hosts entry is almost certainly never correct, though one can "workaround" this by adding a host header to curl, i.e. curl ... -H 'Host: kubernetes.docker.internal' ...
. But the correct solution is to simply add correct host entries, i.e. - mysite.mycompany.com
etc. Hosts in this case are like vhosts in Apache; they're an fqdn that resolves to something the mesh and cluster can use to send requests to. host
, however, in virtualservice
destination
is the service, which is a bit convoluted and is what threw me.
Upvotes: 2