I'm trying to create a custom project template in OpenShift Origin. The Service configuration specifically, looks like below:
{
"kind": "Service",
"apiVersion": "v1",
"metadata": {
"name": "${NAME}",
"annotations": {
"description": "Exposes and load balances the node.js application pods"
}
},
"spec": {
"ports": [
{
"name": "web",
"port": "${APPLICATION_PORT}",
"targetPort": "${APPLICATION_PORT}",
"protocol": "TCP"
}
],
"selector": {
"name": "${NAME}"
}
}
},
where, APPLICATION_PORT is supplied as a user parameter:
"parameters": [
{
"name": "APPLICATION_PORT",
"displayName": "Application Port",
"description": "The exposed port that will route to the node.js application",
"value": "8000"
},
When I try to use this template to create a project, I get the following error:
spec.ports[0].targetPort: Invalid value: "8000": must be an IANA_SVC_NAME (at most 15 characters, matching regex [a-z0-9]([a-z0-9-]*[a-z0-9])*...
I get a similar error in my DeploymentConfig as well, for the http ports in the liveness and readiness probes:
"readinessProbe": {
"timeoutSeconds": 3,
"initialDelaySeconds": 3,
"httpGet": {
"path": "/Info",
"port": "${APPLICATION_ADMIN_PORT}"
}
},
"livenessProbe": {
"timeoutSeconds": 3,
"initialDelaySeconds": 30,
"httpGet": {
"path": "/Info",
"port": "${APPLICATION_ADMIN_PORT}"
}
},
where, APPLICATION_ADMIN_PORT, again, is user-supplied.
Error:
spec.template.spec.containers[0].livenessProbe.httpGet.port: Invalid value: "8001": must be an IANA_SVC_NAME...
spec.template.spec.containers[0].readinessProbe.httpGet.port: Invalid value: "8001": must be an IANA_SVC_NAME...
I've been following https://blog.openshift.com/part-2-creating-a-template-a-technical-walkthrough/ to understand templates, and it, unfortunately, does not have any examples of ports being parameterized anywhere.
It almost seems as if strings are not allowed as the values of these ports. Is that the case? What's the right way to parameterize these values? Should I switch to YAML?
Versions:
OpenShift Master: v1.1.6-3-g9c5694f
Kubernetes Master: v1.2.0-36-g4a3f9c5
Edit 1: I tried the same configuration in YAML format, and got the same error. So, JSON vs YAML is not the issue.
Unfortunately it is not currently possible to parameterize non-string field values: https://docs.openshift.org/latest/dev_guide/templates.html#writing-parameters
" Parameters can be referenced by placing values in the form "${PARAMETER_NAME}" in place of any string field in the template."
Templates are in the process of being upstreamed to Kubernetes and this limitation is being addressed there:
https://github.com/kubernetes/kubernetes/blob/master/docs/proposals/templates.md
The proposal is being implemented in PRs 25622 and 25293 in the kubernetes repo.
edit:
Templates now support non-string parameters as documented here: https://docs.openshift.org/latest/dev_guide/templates.html#writing-parameters
I don't know if this option was available in 2016 when this post was added but now you can use ${{PARAMETER_NAME}} to parameterize non-string field values.
spec:
externalTrafficPolicy: Cluster
ports:
- name: ${NAME}-port
port: ${{PORT_PARAMETER}}
protocol: TCP
targetPort: ${{PORT_PARAMETER}}
sessionAffinity: None
This may a be a bad practice but I'm using sed to substitute int parameters:
cat template.yaml | sed -e 's/PORT/8080/g' > proxy-template-subst.yaml
Template:
apiVersion: template.openshift.io/v1
kind: Template
objects:
- apiVersion: v1
kind: Service
metadata:
name: ${NAME}
namespace: ${NAMESPACE}
spec:
externalTrafficPolicy: Cluster
ports:
- name: ${NAME}-port
port: PORT
protocol: TCP
targetPort: PORT
sessionAffinity: None
type: NodePort
status:
loadBalancer: {}
parameters:
- description: Desired service name
name: NAME
required: true
value: need_real_value_here
- description: IP adress
name: IP
required: true
value: need_real_value_here
- description: namespace where to deploy
name: NAMESPACE
required: true
value: need_real_value_here
Related
I have nextcloud running on bare metal 2 nodes:
node1: 192.168.1.10
node2: 192.168.1.11
In the consul I have defined nextcloud service as such on both the nodes:
{
"service": {
"name": "nextcloud",
"tags": ["nextcloud", "traefik"],
"port": 80,
"check": {
"tcp": "localhost:80",
"args": ["ping", "-c1", "127.0.0.1"],
"interval": "10s",
"status": "passing",
"success_before_passing": 3,
"failures_before_critical": 3
}
}
now this shows up in consul fine:
static config: traefik.yaml
global:
# Send anonymous usage data
sendAnonymousUsage: true
api:
dashboard: true
debug: true
log:
level: DEBUG
entryPoints:
http:
address: ":80"
https:
address: ":443"
serversTransport:
insecureSkipVerify: true
providers:
docker:
endpoint: "unix:///var/run/docker.sock"
exposedByDefault: false
file:
directory: "/config/"
watch: true
consulCatalog:
defaultRule: "Host(`{{ .Name }}.sub.mydomain.com`)"
endpoint:
address: http://127.0.0.1:8500
certificatesResolvers:
linode:
acme:
caServer: https://acme-staging-v02.api.letsencrypt.org/directory
email: myemail#domain.com
storage: acme.json
dnsChallenge:
provider: linode
resolvers:
- "1.1.1.1:53"
- "1.0.0.1:53"
and then dynamic /config/config.yaml:
http:
routers:
nextcloud#consulCatalog:
entryPoints:
- "https"
rule: "Host(`home.sub.mydomain.com`) && Path(`/nextcloud`)"
tls:
certResolver: linode
service: nextcloud
services:
nextcloud:
loadBalancer:
servers:
- url: http://192.168.1.10
- url: http://192.168.1.11
passHostHeader: true
but this shows up as file provider with TLS in instead in addtion to exisiting consulcatalog provider.
and not IP or domain mapped.
actual consulcatalog provider showing up but no tls
I am wondering why my dynamic configuration in http did not updated the nextcloud#consulcatalog and set the https entrypoint.
Any help will be greatly appreciated, I am struggling very hard to get this to work.
I have tried following the docs on traefik but its very confusing specially on the consulcatalog part.
Your configuration is showing up as being defined via the file provider because you are statically defining it in the file at /config/config.yaml.
In order to dynamically retrieve this configuration from Consul, you should not be defining the static config file and instead configure tags on the Consul service registrations that will instruct Traefik to route traffic to your service.
For example:
{
"service": {
"name": "nextcloud",
"tags": [
"nextcloud",
"traefik.enable=true",
"traefik.http.routers.nextcloud.entrypoints=https",
"traefik.http.routers.nextcloud.rule=(Host(`home.sub.mydomain.com`) && Path(`/nextcloud`))",
"traefik.http.routers.nextcloud.tls.certresolver=linode",
"traefik.http.services.nextcloud.loadbalancer.passhostheader=true"
],
"port": 80,
"check": {
"tcp": "localhost:80",
"args": [
"ping",
"-c1",
"127.0.0.1"
],
"interval": "10s",
"status": "passing",
"success_before_passing": 3,
"failures_before_critical": 3
}
}
}
More info can be found on the Routing Configuration docs for Traffic's Consul catalog provider.
Given the following array:
// use strings only in the form <protocol>-<port>
ports: [
'tcp-1514',
'tcp-8080',
'tcp-8443',
],
I'm trying to write jsonnet to split each element of the array to generate this object (represented here in yaml):
ports:
- name: "tcp-1514"
containerPort: 1514
protocol: "tcp"
- name: "tcp-8080"
containerPort: 8080
protocol: "tcp"
- name: "tcp-8443"
containerPort: 8443
protocol: "tcp"
I've tried several iterations of array comprehension to do this, mind you I'm brand new to jsonnet. The latest iteration was something like:
ports: [
{
local proto, port ::= std.split(port_obj, '-');
name: port_obj,
containerPort: port,
protocol: proto,
} for port_obj in $.sharedConfig.ports,
]
where $.sharedConfig.ports is the ports assignment. The problem is local proto, port ::= std.split(port_obj, '-');. I'm not sure this is valid code. The interpreter is poopooing it and I can't find any examples or documentation showing that this is valid.
Ultimately, if it's not valid then I'll have to split() twice, but that would be unfortunate. For instance, this works:
{
local ports = ['tcp-1514', 'tcp-8080', 'tcp-8443',],
ports: [
local port = std.split(name,'-')[1];
local proto = std.split(name,'-')[0];
{
name: name,
protocol: proto,
containerPort: port,
}
for name in ports],
}
which yields:
{
"ports": [
{
"containerPort": "1514",
"name": "tcp-1514",
"protocol": "tcp"
},
{
"containerPort": "8080",
"name": "tcp-8080",
"protocol": "tcp"
},
{
"containerPort": "8443",
"name": "tcp-8443",
"protocol": "tcp"
}
]
}
and YAML:
---
ports:
- containerPort: '1514'
name: tcp-1514
protocol: tcp
- containerPort: '8080'
name: tcp-8080
protocol: tcp
- containerPort: '8443'
name: tcp-8443
protocol: tcp
...but I really dislike the two-line variable assignment. The more I've tested this, the more I believe I'm right determining that the single-line assignment is not doable.
Anyone able to show me how I'm wrong, I'd truly appreciate it.
It may look like a simple answer (that you may have already considered), but well here it goes: using a single local var to hold the split() result, then refer to it in fields' assignments ->
Simple answer:
{
local ports = ['tcp-1514', 'tcp-8080', 'tcp-8443'],
ports: [
local name_split = std.split(name, '-');
{
name: name,
protocol: name_split[0],
containerPort: name_split[1],
}
for name in ports
],
}
Obfuscated answer (no interim local w/split() result):
// Return a map from zipping arr0 (keys) and arr1 (values)
local zipArrays(arr0, arr1) = std.foldl(
// Merge each (per-field) object into a single obj
function(x, y) x + y,
// create per-field object, e.g. { name: <name> },
std.mapWithIndex(function(i, x) { [arr0[i]]: x }, arr1),
{},
);
{
local ports = ['tcp-1514', 'tcp-8080', 'tcp-8443'],
// Carefully ordered set of fields to "match" against: [name] + std.split(...)
local vars = ['name', 'protocol', 'containerPort'],
ports: [
zipArrays(vars, [name] + std.split(name, '-'))
for name in ports
],
}
When I run my API gateway in docker container then it is not able to find my services which are registered in eureka.
API Gateway
-- ocelot.json
{
"ReRoutes": [
{
"DownstreamPathTemplate": "/api/values",
"DownstreamScheme": "http",
"UseServiceDiscovery": true,
"ServiceName": "sampleservice",
"UpstreamPathTemplate": "/sample-api/{catchAll}"
}
],
"GlobalConfiguration": {
"UseServiceDiscovery": true,
"ServiceDiscoveryProvider": {
"Type": "Eureka",
"Host": "myeurekaserver",
"Port": "8761"
}
}
}
-- appsettings.json for API Gateway
{
"eureka": {
"client": {
"shouldRegisterWithEureka": false,
"serviceUrl": "http://myeurekaserver:8761/eureka/",
"ValidateCertificates": false
},
"instance": {
"appName": "gateway",
"hostName": "myeurekaserver",
"port": "7000"
}
}
}
Service Configuration --appsettings.json
{
"eureka": {
"client": {
"shouldRegisterWithEureka": true,
"serviceUrl": "http://myeurekaserver:8761/eureka/",
"ValidateCertificates": false
},
"instance": {
"appName": "sampleservice",
"hostName": "myeurekaserver",
"port": "7001"
}
}
}
docker-compose.yml
version: '3.4'
services:
sampleapi:
image: ${DOCKER_REGISTRY-}sampleapi
ports:
- "7001:80"
networks:
- ecnetwork
build:
context: .
dockerfile: SampleAPI/Dockerfile
gateway:
image: ${DOCKER_REGISTRY-}gateway
ports:
- "7000:80"
networks:
- ecnetwork
build:
context: .
dockerfile: Gateway/Dockerfile
myeurekaserver:
image: ${DOCKER_REGISTRY-}myeurekaserver
ports:
- "8761:8761"
networks:
- ecnetwork
build:
context: .
dockerfile: MyEurekaServer/Dockerfile
networks:
ecnetwork:
external: true
When I run command docker-compose up and check on http://localhost:8761/ I find my services have been registred in the eureka server, but I run http://localhost:7000/sample-api/order
It returns
localhost is currently unable to handle this request. HTTP ERROR 500
I checked my console window, then It is API gateway is able to discover the services, here is the log.
gateway_1 | dbug: Steeltoe.Discovery.Eureka.DiscoveryClient[0]
gateway_1 | FetchRegistryDelta returned: OK
gateway_1 | dbug: Steeltoe.Discovery.Eureka.DiscoveryClient[0]
gateway_1 | FetchRegistry succeeded
It's an application error, check your app API gateway.
500 Internal Server Error
A generic error message, given when an unexpected condition was encountered and no more specific message is suitable
Try to debug your application without Docker.
Check in the docker on which port the service is registered 7000 or 80?
Then see if the 7000 port is accessible for you in local by telnet
when I'm tring to mix the spring cloud config with istio 1.1.1, When my app container(with istio envoy auto-injected) starts, the spring cloud config client will try to get config(applicationContext.yaml) from remote cloud config server(started in advance with good status), unfornately it fails with HTTP 404 error. Even if I've configged my app to have retry for cloud config client, it keeps retring alway with HTTP 404 error(I've confirmed the config server URL is correct from another container) and there's no chance to recover. It happens sometimes. I knew that Istio envoy and my app are in the same kubernetes POD, the app may start before istio envoy, in which case there might be network error but as soon as the envoy is up, everything should be OK. I really don't understand why my app cannot recover automatically. Here're my diagnostic steps:
1. Add retry mechanism in my app(with retry libs included in POM and modified yaml. - retry works but each retry failed with HTTP 404 error
spring-config/
fail-fast: true
retry:
initial-interval: 10000
max-attempts: 100
2. Add 'sleep xx' before my java app starts in my app k8s deployment file - less chance to have HTTP 404 error, but problem is not eliminated
command: ["/bin/sh","-c","sleep 20; java -jar -Xms512m -Xmx1024m app.jar"]
3. get the istio envoy's access log and compare the victim app's and good app's - it sounds like the good log has values for upstream_cluster and upstream_cluster key; the fields for the bad log are empty
the good access log
{
"response_code": "200",
"user_agent": "Java/1.8.0_121",
"response_flags": "-",
"start_time": "2019-06-25T01:17:29.661Z",
"method": "2019-06-25T01:17:29.661Z",
"request_id": "d3d27512-161b-4303-bb48-05a6e19e05b7",
"upstream_host": "172.20.3.104:9083",
"x_forwarded_for": "-",
"requested_server_name": "-",
"bytes_received": "0",
"istio_policy_status": "-",
"bytes_sent": "1144",
"upstream_cluster": "outbound|9083||fota-spring-config.ns-fota.svc.cluster.local",
"downstream_remote_address": "172.20.2.115:45816",
"path": "/fota-spring-config/fota-task/dev/master",
"authority": "fota-spring-config.ns-fota.svc.cluster.local:9083",
"protocol": "HTTP/1.1",
"upstream_service_time": "289",
"upstream_local_address": "-",
"duration": "290",
"downstream_local_address": "172.21.1.152:9083"
}
the bad access log:
{
"upstream_cluster": "-",
"downstream_remote_address": "172.20.2.118:41980",
"path": "/fota-spring-config/fota-dmserver/dev/master",
"authority": "fota-spring-config.ns-fota.svc.cluster.local:9083",
"protocol": "HTTP/1.1",
"upstream_service_time": "-",
"upstream_local_address": "-",
"duration": "0",
"downstream_local_address": "172.21.1.152:9083",
"response_code": "404",
"user_agent": "Java/1.8.0_121",
"response_flags": "NR",
"start_time": "2019-06-25T01:21:24.197Z",
"method": "2019-06-25T01:21:24.197Z",
"request_id": "346716e4-1def-465f-b370-cb1e71e30d25",
"upstream_host": "-",
"x_forwarded_for": "-",
"requested_server_name": "-",
"bytes_received": "0",
"istio_policy_status": "-",
"bytes_sent": "0"
}
the K8S deployment file is attached.
apiVersion: extensions/v1beta1
kind: Deployment
metadata:
name: fota-car
spec:
template:
metadata:
labels:
app: fota-car
version: v1
spec:
serviceAccountName: fota-serviceaccount
imagePullSecrets:
- name: uaes-docker2
containers:
- name: fota-car
image: 192.168.119.22:18080/uaes-fota/fota-car:dev-release-1.0.0
imagePullPolicy: Always
ports:
- containerPort: 8085
env:
- name: SPRING_DATASOURCE_URL
value: jdbc:mysql://mysql-ali-dev.ns-fota-ext-svc/fota-car?useUnicode=true&characterEncoding=utf-8&useSSL=false
- name: SPRING_DATASOURCE_USERNAME
valueFrom:
secretKeyRef:
name: mysql-ali-dev-secret
key: username
- name: SPRING_DATASOURCE_PASSWORD
valueFrom:
secretKeyRef:
name: mysql-ali-dev-secret
key: password
command: ["/bin/sh","-c","java -jar -Xms512m -Xmx1024m app.jar"]
readinessProbe:
httpGet:
path: /actuator/health
port: 18085
initialDelaySeconds: 60
timeoutSeconds: 1
kind: Service
apiVersion: v1
metadata:
labels:
app: fota-car
name: fota-car
spec:
ports:
- name: http
port: 8085
selector:
app: fota-car
I'm trying to proxy an S3 bucket configured as a website from an API Gateway endpoint. I configured an endpoint successfully using the console, but I am unable to recreate the configuration using Cloudformation.
After lots of trial and error and guessing, I've come up with the following CF stack template that gets me pretty close:
Resources:
Api:
Type: 'AWS::ApiGateway::RestApi'
Properties:
Name: ApiDocs
Resource:
Type: 'AWS::ApiGateway::Resource'
Properties:
ParentId: !GetAtt Api.RootResourceId
RestApiId: !Ref Api
PathPart: '{proxy+}'
RootMethod:
Type: 'AWS::ApiGateway::Method'
Properties:
HttpMethod: ANY
ResourceId: !GetAtt Api.RootResourceId
RestApiId: !Ref Api
AuthorizationType: NONE
Integration:
IntegrationHttpMethod: ANY
Type: HTTP_PROXY
Uri: 'http://my-bucket.s3-website-${AWS::Region}.amazonaws.com/'
PassthroughBehavior: WHEN_NO_MATCH
IntegrationResponses:
- StatusCode: 200
ProxyMethod:
Type: 'AWS::ApiGateway::Method'
Properties:
HttpMethod: ANY
ResourceId: !Ref Resource
RestApiId: !Ref Api
AuthorizationType: NONE
RequestParameters:
method.request.path.proxy: true
Integration:
CacheKeyParameters:
- 'method.request.path.proxy'
RequestParameters:
integration.request.path.proxy: 'method.request.path.proxy'
IntegrationHttpMethod: ANY
Type: HTTP_PROXY
Uri: 'http://my-bucket.s3-website-${AWS::Region}.amazonaws.com/{proxy}'
PassthroughBehavior: WHEN_NO_MATCH
IntegrationResponses:
- StatusCode: 200
Deployment:
DependsOn:
- RootMethod
- ProxyMethod
Type: 'AWS::ApiGateway::Deployment'
Properties:
RestApiId: !Ref Api
StageName: dev
Using this template I can successfully get the root of the bucket website, but the proxy resource gives me a 500:
curl -i https://abcdef.execute-api.eu-west-1.amazonaws.com/dev/index.html
HTTP/1.1 500 Internal Server Error
Content-Type: application/json
Content-Length: 36
Connection: keep-alive
Date: Mon, 11 Dec 2017 16:36:02 GMT
x-amzn-RequestId: 6014a809-de91-11e7-95e4-dda6e24d156a
X-Cache: Error from cloudfront
Via: 1.1 8f6f9aba914cc74bcbbf3c57e10df26a.cloudfront.net (CloudFront)
X-Amz-Cf-Id: TlOCX3eemHfY0aiVk9MLCp4qFzUEn5I0QUTIPkh14o6-nh7YAfUn5Q==
{"message": "Internal server error"}
I have no idea how to debug that 500.
To track down what may be wrong, I've compared the output of aws apigateway get-resource on the resource I created manually in the console (which is working) with the one Cloudformation made (which isn't). The resources look exactly alike. The output of get-method however, is subtly different, and I'm not sure it's possible to make them exactly the same using Cloudformation.
Working method configuration:
{
"apiKeyRequired": false,
"httpMethod": "ANY",
"methodIntegration": {
"integrationResponses": {
"200": {
"responseTemplates": {
"application/json": null
},
"statusCode": "200"
}
},
"passthroughBehavior": "WHEN_NO_MATCH",
"cacheKeyParameters": [
"method.request.path.proxy"
],
"requestParameters": {
"integration.request.path.proxy": "method.request.path.proxy"
},
"uri": "http://muybucket.s3-website-eu-west-1.amazonaws.com/{proxy}",
"httpMethod": "ANY",
"cacheNamespace": "abcdefg",
"type": "HTTP_PROXY"
},
"requestParameters": {
"method.request.path.proxy": true
},
"authorizationType": "NONE"
}
Configuration that doesn't work:
{
"apiKeyRequired": false,
"httpMethod": "ANY",
"methodIntegration": {
"integrationResponses": {
"200": {
"responseParameters": {},
"responseTemplates": {},
"statusCode": "200"
}
},
"passthroughBehavior": "WHEN_NO_MATCH",
"cacheKeyParameters": [
"method.request.path.proxy"
],
"requestParameters": {
"integration.request.path.proxy": "method.request.path.proxy"
},
"uri": "http://mybucket.s3-website-eu-west-1.amazonaws.com/{proxy}",
"httpMethod": "ANY",
"requestTemplates": {},
"cacheNamespace": "abcdef",
"type": "HTTP_PROXY"
},
"requestParameters": {
"method.request.path.proxy": true
},
"requestModels": {},
"authorizationType": "NONE"
}
The differences:
The working configuration has responseTemplates set to "application/json": null. As far as I can tell, there's no way to set a mapping explicitly to null using Cloudformation. My CF method instead just has an empty object here.
My CF method has "responseParameters": {},, while the working configuration does not have responseParameters at all
My CF method has "requestModels": {},, while the working configuration does not have requestModels at all
Comparing the two in the console, they are seemingly exactly the same.
I'm at my wits end here: what am I doing wrong? Is this possible to achieve using Cloudformation?
Answer: The above is correct. I had arrived at this solution through a series of steps, and re-applied the template over and over. Deleting the stack and deploying it anew with this configuration had the desired effect.