## Docker in MyScript WebServices
Mathieu Ruellan - https://github.com/mathieuruellan
## Activity
We are selling technology around handwriting recognition:
- Text
- Maths
- Shapes
- Music
[Visit our dev-portal: https://dev.myscript.com/](https://dev.myscript.com/ "Visit dev-portal!")
## Products
- SDK: C, java, .net
- Widgets: IOS, Android (and soon windows 10)
- Services: Recognition with REST or WebSockets
[Visit our dev-portal: https://dev.myscript.com/](https://dev.myscript.com/ "Visit dev-portal!")
## What we do with Docker?
- MyScript Cloud (online recognition)
- MyScript ATK (certificate provider for mobile apps embedding our widgets)
- Builds
## Who is using Docker?
- developers
- testers
- production (AWS/EC2 environment)
## What do we need in production?
- Backend
- DB (mongodb)
- SMTP relay (Postfix)
- HAProxy (load balancing)
- logger (rsyslog - elasticsearch/logstash/kibana)
- Varnish
- static content (nginx)
## Deliveries are Docker Images
- Environment agnostic (thanks to environment variables)
- Build results are pushed in a local registry (it's also a docker container)
- Production deployment on private S3 with **`docker save`** and **`docker load`**
- Possibility to run a production-like environment with fig/docker-compose
## Best practices
- One process per container
- Pipe the process stdout/stderr to logger otherwise mount a /dev/log volume (deprecated with --log-driver since 1.6)
- In Production, use **`docker run --net=host ...`**
## **Do not believe it's secured!!!**
## Docker and test
It's light, fast and everything is scriptable!
## Example: Mailing testing
## Faking AWS services 1/2
- Creating a smtp/pop3 container
- Creating AWS/ELB fake with HAProxy
## Faking AWS services 2/2
## Automatic tests with a Makefile
```
_waittcp:
@echo -n "Waiting for TCP Port on IP:$(WAIT_IP):$(WAIT_PORT) " && \
CPT=0 && \
while ! nc -w 1 $(WAIT_IP) $(WAIT_PORT); do \
((CPT++)); \
if [ $$CPT -gt 60 ]; then \
echo "timeout"; \
exit 1; \
fi; \
sleep 1; \
echo -n .; \
done; \
echo " Found!"
_start-mailtest:
@echo -n "starting mailtest container. " && \
docker run $(DOCKER_MAILTEST_PARAMETERS) -d -h test.myscript.com -e maildomain=myscript.com -e user=qawebservices:qawebservices,test1:test1,test2:test2,test3:test3 --name $(TEST_DOCKER_MAILTEST_INSTANCE_NAME) webtest:5000/myscript-webservices-mailtest:$(DOCKERTAG) && \
MAILTESTIP=$$(docker inspect --format '{{ .NetworkSettings.IPAddress }}' $(TEST_DOCKER_MAILTEST_INSTANCE_NAME)) && \
$(MAKE) WAIT_IP=$${MAILTESTIP} WAIT_PORT=110 _waittcp && \
$(MAKE) WAIT_IP=$${MAILTESTIP} WAIT_PORT=25 _waittcp
_start-license-volumes:
@echo -n "Starting license-7.0 volume. " && docker run -d --name $(TEST_DOCKER_LICENSE_7_0_INSTANCE_NAME) webtest:5000/myscript-cloud-volume-licensekey-7.0:$(DOCKERTAG)
_start-mongo-docker:
@echo -n "Starting mongo container. " && docker run $(DOCKER_MONGO_PARAMETERS) -d -h mongo --name $(TEST_DOCKER_MONGO_INSTANCE_NAME) mongo:2.6 --smallfiles --replSet "testreplset"
@MONGOIP=$$(docker inspect --format '{{ .NetworkSettings.IPAddress }}' $(TEST_DOCKER_MONGO_INSTANCE_NAME)) && \
$(MAKE) WAIT_IP=$${MONGOIP} WAIT_PORT=27017 _waittcp && \
echo "Initializing ReplicaSet." && \
docker exec $(TEST_DOCKER_MONGO_INSTANCE_NAME) mongo --quiet --eval "rs.initiate();" > /dev/null
_start-backend-docker:
@echo -n "Starting current version backend container. " && \
MAILTESTIP=$$(docker inspect --format '{{ .NetworkSettings.IPAddress }}' $(TEST_DOCKER_MAILTEST_INSTANCE_NAME)) && \
docker run \
$(DOCKER_BACKEND_PARAMETERS) \
-d \
--volumes-from $(TEST_DOCKER_LICENSE_7_0_INSTANCE_NAME) \
--link $(TEST_DOCKER_MONGO_INSTANCE_NAME):mongo \
-e "MYSCRIPT_CLOUD_LOGGER_HOST=" \
-e "MYSCRIPT_CLOUD_SMTP_SERVER=$${MAILTESTIP}" \
-e "MYSCRIPT_CLOUD_SEND_EMAIL_TO_USER=false" \
-e "MYSCRIPT_CLOUD_ACTIVATION_EMAIL_QAWEBERVICES=qawebservices@myscript.com" \
--name $(TEST_DOCKER_BACKEND_INSTANCE_NAME) \
webtest:5000/myscript-cloud-backend:$(DOCKERTAG)
@BACKENDIP=$$(docker inspect --format '{{ .NetworkSettings.IPAddress }}' $(TEST_DOCKER_BACKEND_INSTANCE_NAME)) && \
$(MAKE) WAIT_IP=$${BACKENDIP} WAIT_PORT=8894 _waittcp
```
## Builds
- The jenkins is deployed in a docker container!
- Everybody can test the jenkins on his workstation!
## And What about the production?
Only one AMI! The EC2 AMI is a stable debian with:
- docker.io
- jq
- awscli
EC2 instances run a script injected through user-data, configured with tags
- clean containers and images
- docker load from S3 repository
- launch dockers with environment variables
## Monitoring
User-data scripts create cron jobs to
- check host states (disk, cpu etc.)
- docker containers status
- launch periodic backups (it's also a docker)
## Difficulties
- Docker comes with a DevOps approach. Changing actual practices
- Workstation must be linux
- New tool to learn (testers have to get IT knowledge)
- Tools to help them
- Can't migrate everything in one step, still working with the past
## Benefits
- Force people to have good practices
- provide something testable
- no more 'preprod' for testers
- Update deployment facilities
- Less stress for IT
