Hi Juan, ... thanks for chiming in... and thanks for the pointer to the Git repository with the Dockerfiles; that makes life already a lot easier.
You are right, once you have the docker-compose.yml files lined up it's not a big stretch to setup a Swarm cluster. After running such a cluster for a client of mine I just want to add: - a Swarm cluster with less than 3 nodes will not run very stable; I'm saying this, because at the moment we have 2 servers at our disposal (I think) - colleagues told me that Swarm cluster worked less reliable for them than other solutions; I don't think for a demo system that is too much of a concern, but again I had one running in a production environment and had no major problems If you want then let's connect ([email protected]) and figure out how we can proceed with this... before the next GSoC season begins ;-) Cheers On Sat, Jan 19, 2019 at 4:03 PM Juhan Aasaru <[email protected]> wrote: > Hi Aleks > > Thanks for your work on pushing the demo server to live. > I have played around with the containers also and I add my feedback and > ideas. > > > - it would come in handy to have default Fineract CN Docker images > published on Docker Hub > > I think this is a way to go. If we want to promote adoption of Fineract-CN > then public images > lower the burden to anyone to download and get going with the project. > Does the CI server already exist that could potentially build the images? > > > - I suggest to add a Dockerfile in every microservice Git repository > (e. > > g. fineract-cn-customer, fineract-cn-teller, fineract-cn-payroll) and > let a > > CI server build and publish Docker images of these > > Yes. Most of the Dockerfiles already exist here > https://github.com/openMF/fineract-cn-containers > But they logically belong to the application's own code base so I see no > harm in adding all > Dockerfiles to the app's own repository. > > - I am assuming that we **don't** want to go all the way to setup a > Kubernetes (or even a Docker Swarm) cluster; the goal is to just have a set > > If we plan to operate with docker-compose already (and run in two servers) > then I, in my opinion, it wouldn't be much overhead to create a Swarm > cluster. > If I look at the instructions (https://docs.docker.com/get-started/part4/) > it doesn't seem like a lot of work. > Also if something happens then Swarm can detect and relaunch containers. > But I'm no system administrator myself so I might be mistaken in terms of > how much work it requires. > > > - to avoid code changes or Docker image rebuilds we should introduce > > environment variables in the application.yml files of these > microservice > > projects; e. g.: > > cassandra: > > clusterName: staging_cluster > /---/ > > ... should look something like this: > > cassandra: > > clusterName: > ${FINERACT_CUSTOMER_CASSANDRA_CLUSTER_NAME:staging_cluster} > /---/ > > I think there is no need to change application.yml files. > In docker-compose.yml you can overwrite any application.yml property in > "environment" section like this: > > environment: > - > > "cassandra.clusterName=${FINERACT_CUSTOMER_CASSANDRA_CLUSTER_NAME:staging_cluster}" > > Juhan > > > > Kontakt Aleksandar Vidakovic (<[email protected]>) kirjutas > kuupäeval N, 17. jaanuar 2019 kell 03:46: > > > Hi everyone, > > > > ... hope you all enjoyed the holidays and had a good start into the new > > year :-) > > > > I have to appologize for my radio silence concerning the demo server, > but I > > got a bit steam rolled by work in the last 6 months. > > > > Anyway, I just wanted to get this effort going again and would like to > > discuss it with anyone interested. > > > > The current status: > > > > - we have 2 (quite big) servers provided by the Apache Foundation to > run > > the demo setup > > - initially I tried to get it running on one, but was not enough (even > > with 32GB of RAM and some swap configuration tricks) > > - I've used the demo server module with some minor modifications to > turn > > off non-essential components (thanks Myrle) > > > > Trying all of this took quite some time... even on the beefy machine from > > Apache it took (as far as I remember) 30-40min until the demo server > > startup would ultimately fail. > > > > Instead of going down that route again I'd like to propose a different > > strategy: > > > > - I am assuming that we **don't** want to go all the way to setup a > > Kubernetes (or even a Docker Swarm) cluster; the goal is to just have > a > > set > > of docker-compose.yml files to start the system > > - it would come in handy to have default Fineract CN Docker images > > published on Docker Hub > > - I suggest to add a Dockerfile in every microservice Git repository > (e. > > g. fineract-cn-customer, fineract-cn-teller, fineract-cn-payroll) and > > let a > > CI server build and publish Docker images of these > > - to avoid code changes or Docker image rebuilds we should introduce > > environment variables in the application.yml files of these > microservice > > projects; e. g.: > > > > [code] > > ... > > cassandra: > > clusterName: staging_cluster > > contactPoints: 127.0.0.1:9042,127.0.0.2:9042,127.0.0.3:9042 > > keyspace: seshat > > cl: > > read: LOCAL_QUORUM > > write: LOCAL_QUORUM > > delete: LOCAL_QUORUM > > ... > > [/code] > > > > ... should look something like this: > > > > [code] > > ... > > cassandra: > > clusterName: > ${FINERACT_CUSTOMER_CASSANDRA_CLUSTER_NAME:staging_cluster} > > contactPoints: ${FINERACT_CUSTOMER_CASSANDRA_CONTACT_ENDPOINTS: > > 127.0.0.1:9042,127.0.0.2:9042,127.0.0.3:9042} > > keyspace: ${FINERACT_CUSTOMER_CASSANDRA_KEYSPACE:seshat} > > cl: > > read: LOCAL_QUORUM > > write: LOCAL_QUORUM > > delete: LOCAL_QUORUM > > ... > > [/config] > > > > - with the above changes we could then define docker-compose.yml files > > like this (pseudo file for customer microservice): > > > > [code] > > version: '3.6' > > > > services: > > customer: > > image: nexus.pelotoninnovations.com/rspndr/server-in-memory:latest > > depends_on: > > - mongo > > env_file: > > - ./customer.env > > ports: > > - "10000:10000" > > command: sh -c "java -Xmx1024m -Duser.timezone=UTC > > -Dlogging.config=./logback.xml -jar -Djava.net.preferIPv4Stack=true > > fineract-cn-customer.jar" > > [/code] > > > > ... and the customer.env file would contain something like this: > > > > [code] > > FINERACT_CUSTOMER_CASSANDRA_CLUSTER_NAME=prod_cluster > > > > > FINERACT_CUSTOMER_CASSANDRA_CONTACT_ENDPOINTS=server1:9042,server2:9042,server3:9042 > > FINERACT_CUSTOMER_CASSANDRA_KEYSPACE=seshat > > [/code] > > > > - we would provide templates for those env files (e. g. > > "customer.env.template"); custom configurations (e. g. "customer.env") > > should not be checked into Git > > - if no environment variables are provided then the defaults in the > > application.yml config files kick in with reasonable values for a > local > > dev > > machine deployment (given the required resources unlikely for most > devs) > > > > > > Advantages: > > > > - ready to consume Fineract CN Docker images > > - no lengthy builds > > - no re-build (Gradle, Docker) for configuration changes > > - no requirement to do cluster (Swarm, Kubernetes) setup > > - the Docker images could still be used as the basic building blocks > of > > more complex architectures (Kubernetes) > > - every service can be started/stopped separately which makes life a > lot > > easier when we have to figure out the right configuration for the demo > > setup (I guess it would make it also easier for others that would like > > to > > setup their own environments) > > > > I am using most (if not all) of the required bits and pieces for this > setup > > on a daily basis and I think it should be not too complicated to get this > > working. And it would not interfere (too much) with the existing Git > > repositories. > > > > Please let me know what you think... > > > > Cheers, > > > > Aleks > > >
