I have a docker-compose file with two services: app and httpd
app
app:
image: primus852/machinelearning:latest
ports:
- 5001:5000
expose:
- "5001"
restart: always
networks:
- default
volumes:
- ./api:/app
environment:
- FLASK_APP=app/source/__init__.py
- FLASK_ENV=development
httpd
httpd:
image: primus852/mitswiki:latest
ports:
- 80:80
restart: always
networks:
- default
volumes:
- ./project:/var/www/html
Flask app
The app container has an endpoint like this:
#app.route('/predict', methods=['GET'])
def predict():
...DO STH....
I can open http://localhost:5001/predict in my browser, works...
I can curl from my cmd: curl localhost:5001/predict, works...
But when I am inside my httpd container this does not work from the console: curl localhost:5001/predict
curl: (7) Failed to connect to localhost port 5001: Connection refused
So I thought I address the app container as I address my mysql from inside my httpd container: curl app:5001/predict but it has the same result.
Can anyone see what I am doing wrong?
According to your yaml:
ports:
- 5001:5000
Inside container you have to use port 5000
Inside the httpd container localhost refers to just that httpd container. It cannot access other containers by default.
Another thing which might be occuring is that your app is not open for 'remote' access. A connection from one container to another one is a remote connection.
Within your docker-compose files you can link containers to eachother
While the containers are linked you can then use curl to get the /predict page by using curl app:5001/predict
Related
I am trying to connect MinIO with KeyCloak and I follow the instructions provided in this documentation:
https://github.com/minio/minio/blob/master/docs/sts/keycloak.md
What I have done so far is deploy a Docker container for the MinIO server, another one for the MinioClient and a third one used for the KeyCloak server.
As you can see in the following snippet the configuration of the Minio Client container is done correctly, since I can list the buckets available in the Minio Server:
mc ls myminio
[2020-05-14 11:54:59 UTC] 0B bucket1/
[2020-05-06 12:23:01 UTC] 0B bucket2/
I have an issue arising when I try to configure MinIO as depicted in step 3 (Configure MinIO) of the documentation. In more detail, the command that I run is this one:
mc admin config set myminio identity_openid config_url="http://localhost:8080/auth/realms/demo/.well-known/openid-configuration" client_id="account"
And the error I get is this one:
mc: <ERROR> Cannot set 'identity_openid config_url=http://localhost:8080/auth/realms/demo/.well-known/openid-configuration client_id=account' to server. Get http://localhost:8080/auth/realms/demo/.well-known/openid-configuration: dial tcp 127.0.0.1:8080: connect: connection refused.
When I curl this address http://localhost:8080/auth/realms/demo/.well-known/openid-configuration from the MinIO Client container though, I retrieve the JSON file.
Turns out, all I had to do is change the localhost in the config_url, from localhost to the IP of the KeyCloak container (172.17.0.3).
This is just a temporary solution that works for now, but I will continue searching for something more concrete than just hardcoding the IP.
When I figure out the solution, this answer will be updated.
Update
I had to create a docker-compose.yml file as the one below in order to overcome the issues without having to manually place the IP of the KeyCloak container.
version: '2'
services:
miniod:
image: minio/minio
restart: always
container_name: miniod
ports:
- 9000:9000
volumes:
- "C:/data:/data"
environment:
- "MINIO_ACCESS_KEY=access_key"
- "MINIO_SECRET_KEY=secret_key"
command: ["server", "/data"]
networks:
- minionw
mcd:
image: minio/mc
container_name: mcd
networks:
- minionw
kcd:
image: quay.io/keycloak/keycloak:10.0.1
container_name: kcd
restart: always
ports:
- 8080:8080
environment:
- "KEYCLOAK_USER=admin"
- "KEYCLOAK_PASSWORD=pass"
networks:
- minionw
networks:
minionw:
driver: "bridge"
Connection refused occurs when a port is not accessible on the hostname or IP we specified.
Please try exposing the port using --expose flag along with the port number which you wish to expose when using the docker CLI. Then being exposed, you can access on it on localhost
I have a docker like this:
version: '3.5'
services:
RedisServerA:
container_name: RedisServerA
image: redis:3.2.11
command: "redis-server --port 26379"
volumes:
- ../docker/redis/RedisServerA:/data
ports:
- 26379:26379
expose:
- 26379
RedisServerB:
container_name: RedisServerB
image: redis:3.2.11
command: "redis-server --port 6379"
volumes:
- ../docker/redis/RedisServerB:/data
ports:
- 6379:6379
expose:
- 6379
Now I do a vagrant ssh and do
ping RedisServerA
ping RedisServerB
They both work.
Now I try to connect to the redis server:
redis-cli -h RedisServerB
Works fine
Then I try to connect to the other
redis-cli -h RedisServerA -p 26739
It says:
Could not connect to Redis at RedisServerA:26739: Connection refused
Could not connect to Redis at RedisServerA:26739: Connection refused
Twice.
What am I missing here?
Typically in this setup you'd let each container run on its "natural" port. For connections from outside Docker you need the ports: mapping, and you'd access a container via its published port on the host's IP address. For connections between Docker containers (assuming they're on the same network, and if you used bare docker run, you manually created that network), you use the container name and the container's internal port number.
We can clean up the docker-compose.yml file by removing some unnecessary lines (container_name: and expose: don't really have a practical effect) and letting the image run its default command: on the default port, and only remapping with ports:. We'd get:
version: '3.5'
services:
RedisServerA:
image: redis:3.2.11
volumes:
- ../docker/redis/RedisServerA:/data
ports:
- 26379:6379
RedisServerB:
image: redis:3.2.11
volumes:
- ../docker/redis/RedisServerB:/data
ports:
- 6379:6379
Between containers, you'd use the default port
redis-cli -h RedisServerA
redis-cli -h RedisServerB
From outside Docker you'd use the server's host name and the published ports
redis-cli -h server.example.com -p 23679
redis-cli -h server.example.com
[https://github.com/gtriggiano/ngrok-tunnel ] runs ngrok inside a container. Ngrok is required to run in the container to avert security risks. But am facing problems after running the scripts, which generates the url
$ docker pull gtriggiano/ngrok-tunnel
$ docker run -it -e "TARGET_HOST=localhost" -e "TARGET_PORT=3000" -p 4040 gtriggiano/ngrok-tunnel
am running my rails app on localhost:3000
is it my problem or can it be fixed by altering the scripts(inside the repo)?
I couldn't get this working but switched to https://github.com/shkoliar/docker-ngrok and it works brilliantly.
In my case I added it to my docker-compose.yml file:
ngrok:
image: shkoliar/ngrok:latest
ports:
- 4551:4551
links:
- web
environment:
- PARAMS=http -region=eu -authtoken=${NGROK_AUTH_TOKEN} localdev.docker:80
networks:
dev_net:
ipv4_address: 10.5.0.10
And it's started with everything else when I do docker-compose up -d
Then there's a web UI at http://localhost:4551/ for you to see the status, requests, the ngrok URLs, etc.
The Github page does have examples of running it manually from the command line too though, rather than via docker-compose:
Command-line Example The example below assumes that you have running
web server docker container named dev_web_1 with exposed port 80.
docker run --rm -it --link dev_web_1 shkoliar/ngrok ngrok http dev_web_1:80
With command line usage, ngrok session is active until it
won't be terminated by Ctrl+C combination.
No. if you execute -p with single number it's container port - host port is randomly assigned.
Using -p, --publish ip:[hostPort]:containerPort at docker run can specify the the host port with the container port.
as of now the 4040 of container is exposed. Not sure if your service listens by default on it.
To get localhost port execute
docker ps
you'll see the actual port it's not listening on.
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
1aaaeffe789d gtriggiano/ngrok-tunnel "npm start" About a minute ago Up About a minute 0.0.0.0:32768->4040/tcp wizardly_poincare
here it's listening on localhost:32768
this composer works for me. Note that in the entrypoint command for ngrok you have to reference the other service by name
version: '3'
services:
yourwebserver:
build:
context: ./
dockerfile: ...
target: ...
container_name: yourwebserver
volumes:
- ...
ports:
- ...
extra_hosts:
- 'host.docker.internal:host-gateway'
depends_on:
- ngrok
ngrok:
image: ngrok/ngrok:alpine
environment:
NGROK_AUTHTOKEN: '...'
command: 'http yourwebserver:80'
ports:
- '4040:4040'
expose:
- '4040'
I'm not sure if you have already solved this but when I was getting this error I could only solve it like this:
# docker-compose.yml
networks:
- development
I also needed to expose the 3000 port of my web container because it still wasn't exposed.
# docker.compose.yml
web:
expose:
- "3000"
My container for the server running on development is also under the development network. The only parameters, I believe, you should pass for the container to execute are image, ports, environment with DOMAIN and PORT for the server container, a link, and an expose on your web container:
# docker-compose.yml
ngrok:
image: shkoliar/ngrok
ports:
- 4551:4551
links:
- web
networks:
- development
environment:
- DOMAIN=squad_web
- PORT=3000
Actually to make ngrok work with your docker container you can install it outside of your project just like the manual on their website says. And then add
nginx:
labels:
- "traefik.http.routers.${PROJECT_NAME}_nginx.rule=Host(`${PROJECT_BASE_URL}`, `aaa-abc-xxx-140-177.eu.ngrok.io`)"
This particular example is for docker4drupal docker-compose file and traefik mapped as 80:80
Context
I was planning on simplifying some development setup of multiple docker-compose.yml by introducing virtual hosts locally. I looked around and decided to use nginx-proxy for the reverse-proxy (ability to set VIRTUAL_HOST for each service).
Setup
To expose these on the host machine I went the route of dnsmasq and adding a /etc/resolver/test/ with nameserver 127.0.0.1.
I went and put the above into action using a dev/docker-compose.yml file:
version: '3.5'
services:
nginx-proxy:
image: jwilder/nginx-proxy
restart: 'always'
ports:
- "80:80"
- "443:443"
volumes:
- "/var/run/docker.sock:/tmp/docker.sock:ro"
dnsmasq:
image: andyshinn/dnsmasq
restart: 'always'
ports:
- "53:53/tcp"
- "53:53/udp"
cap_add:
- NET_ADMIN
command: --log-facility=-
volumes:
- ./data/dnsmasq.conf:/etc/dnsmasq.conf
- ./data/dnsmasq.d:/etc/dnsmasq.d
networks:
default:
external:
name: proxynet
The data/dnsmasq.conf file only contains address=/test/127.0.0.1.
I've also created an external network proxynet and use that as the default network for the docker-compose file(s) (docker network create proxynet). This then allows other docker-compose files and services to be linked to the proxy.
I have the following proj1/docker-compose.yml:
version: "3.5"
services:
proj1-web:
image: jwilder/whoami
environment:
- VIRTUAL_HOST=proj1-web.test
networks:
default:
external:
name: proxynet
Having both these of these docker-compose files running (i.e., docker-compose up) I am able to access proj1-web.test from my local machine. Everything works as expected.
Now I want to be able to reference proj1-web.test in another container and have it resolve to the running container.
I'll create proj2/docker-compose.yml (similar to previous just different name):
version: "3.5"
services:
proj2-web:
image: jwilder/whoami
environment:
- VIRTUAL_HOST=proj2-web.test
networks:
default:
external:
name: proxynet
With everything running I can access both proj1-web.test and proj2-web.test from my local machine. I can successfully curl different services using between proj1 and proj2: docker-compose run proj1-web sh -c "apk update -qq; apk add curl -qq; curl -v proj2-web:8000".
Problem
The problem is that I cannot curl the virtual host's name proj2-web.test from proj1: docker-compose run proj1-web sh -c "apk update -qq; apk add curl -qq; curl -v proj2-web.test":
* Rebuilt URL to: proj2-web.test/
* Trying 127.0.0.1...
* TCP_NODELAY set
* connect to 127.0.0.1 port 80 failed: Connection refused
* Failed to connect to proj2-web.test port 80: Connection refused
* Closing connection 0
curl: (7) Failed to connect to proj2-web.test port 80: Connection refused
Is there something I'm missing here? It appears the individual containers don't have access to the DNS being provided from dnsmasq to my local machine, I cannot figure out how to grant them that access. Maybe I'm going about this the wrong way -- I am open to suggestions.
I ended up creating a solution which addresses my question. You can see the repository here for the tool:
https://github.com/scoremedia/dcdc
I also created a blog post detailing a bit of this: https://kevinjalbert.com/docker-compose-dns-consistency-dcdc/
Hopefully this helps others.
I'm using Docker for Mac. I have two containers.
1st: A PHP application that is attempting to connect to localhost:3306 to MySQL.
2nd: MySQL
When running with links, they are able to reach each other.
However, I would like to avoid changing any of the code in the PHP application (e.g. changing localhost to "mysql") and stay with using localhost.
Host networking seems to do the trick, the problem is, when I enable host networking I can't access the PHP application on port 80 on my host mac.
If I docker exec -it into the php application and curl localhost, i see the HTML, so it looks like the port is just not forwarding to the host machine?
this is an example for docker-compose
it runs mysql in one container and phpmyadmin in another
the containers are linked together
you can access the containers via your host machine on the ports
3316 and 8889
my_mysql:
image: mysql/mysql-server:latest
container_name: my_mysql
environment:
- MYSQL_ROOT_PASSWORD=1234
- MYSQL_DATABASE=test
- MYSQL_USER=test
- MYSQL_PASSWORD=test
ports:
- 3316:3306
restart: always
phpmyadmin:
image: phpmyadmin/phpmyadmin
container_name: my_myadmin
links:
- my_mysql:my_mysql
environment:
- PMA_ARBITRARY=0
- PMA_HOST=my_mysql
ports:
- 8889:80
restart: always