Sandbox Infra requirements

[vvujjini]

As discussed during our calls, Co-Develop is funding a DevOps engineer to help setup and manage AWS account for G2P-Connect sandbox environment. This sandbox environnemnt will help us showcase integrated end to end demos using G2P-Connect interfaces.

Request all the DPG leads to call out infra requirements to this thread. Over next few days, I shall coordinate and share AWS account details with the respective DPG tech leads.

PLEASE NOTE: Co-Develop funded Devops engineer shall ensure AWS infra layer provision and support while respective DPGs tech teams MUST install, configure and integrate respective application layer components.

[pramodkvarma]

@vvujjini you can send the link to this post via email to all participants to encourage them to join the discussion board and respond to these.

[jeremi]

@vvujjini For OpenG2P hosting, we would need an RDS Postgres and a VM like c6a.xlarge to run docker.
Ed or Avik can comment on the PaymentHub part.

[vvujjini]

@jeremi noted. I will have the environment details available in next few days.

[edcable]

I'm discussing this with @Avik and he will reply to this thread shortly.

Ed

[fynmanoj]

m6g.2xlarge for transactional and operational components. Zeebe/Orchestrator, Elasticsearch, MySQL, and Kafka are the main memory utilizing components

[vvujjini]

@fynmanoj Can please call out the DPG name?

[fynmanoj]

Mifos and Fineract

[euanmillar]

Hi, Euan from OpenCRVS here. We require 1 (basic), or 3 (recommended), dedicated Ubuntu servers to run our stack in its Docker Swarm configuration. Specifics and docs for your DevOps engineer are here: https://documentation.opencrvs.org/setup/3.-installation/3.3-set-up-a-server-hosted-environment We have an Ansible playbook to set up the servers. Our deploy script provisions all the software we need such as MongoDB, ElasticSearch, InfluxData. We can discuss amending the deploy script if we want to share resources as long as they are available to the Docker Swarm. It would be easier for our environment to be standalone and we can enable API access to all other systems, such as MOSIP which we have integrated with already.

[euanmillar]

1 t2.2xlarge is fine for us too.

[malike]

Hi, CORE MIS here. We will require a t2.2xlarge. This will suffice for our use case deployment.

[tdaly61]

Hi folks Tom Daly from the Mojaloop community here. Mojaloop runs entirely in kubernetes and employs helm charts to package and deploy all the services. We have single node installation automated such that we can have a running Mojaloop switch on an t2.2xlarge vm in < 1 hr so we just require that shape i.e. t2.2xlarge and perhaps allow 200GB root disk for convenience in an Ubuntu 20 instance. I state it this way because I am assuming we don't need to scale to large TPS for this demo.

[vvujjini]

@tdaly61 Yes. The plan is to get the base machines available along with required keys/access info for teams to install applications/services by tomorrow. There may be some minor teething issues in first few days. DevOps team Santosh / Geetha will help resolve these at core infra level 1:1 with teams.

Considering some sensitivity with credentials, I've created 1:1 emails with DPGs to coordinate as an exception! Otherwise, these discussion forums will be preferred for everyone's benefit too. Thanks.

[tdaly61]

hi @pramodkvarma , do you have any updates to share on the provisioning ? ...thanks much.

[pramodkvarma]

@vvujjini can you Santhosh get the provisioning going asap? Thanks.

[gandham-santhosh]

@tdaly61 we will have infra provisioning done today and will share the access details over email ASAP

[tdaly61]

@gandham-santhosh : thanks much!

[tdaly61]

Hi folks , thanks so much for setting up the hardware, I have our **demo ** instance of Mojaloop v14 running in the sandbox VM. So now my next question is how can the Mifos folks access the (http) endpoints and are there any externally accessible i.e. public IPs to use or are we going to be doing this all on internal AWS vpc's ? thanks again.

[gandham-santhosh]

@tdaly61 We will create a subdomain mojaloop.g2pconnect.org and update the DNS pointing to the Mojaloop EC2 instance so that Mifos team can access it via domain name. Hope this works with you

[tdaly61]

@gandham-santhosh : Humm, is there any reason you couldn't add public IPs and allow ingress on port 80 and 443 ? Reason for this is that while a private subnet will allow Mifos <-> ML I think the Mifos folks like us at ML will have some major inconveniences trying to integrate and test things without some form of browser access. ? If you are reluctant to do that, what about if we limit the ingress IPs for the port 80 and 443 access ?

[gandham-santhosh]

@tdaly61 As part of mapping the domain name to the Mojaloop instance, we will be creating an Elastic Public IP, assigning it to Mojaloop instance, and updating the DNS A record. After this is done, Mifos team will be able to access Mojaloop via domain name through the browser or API Call. Do you see any challenges with it?

[tdaly61]

@gandham-santhosh : ahh that sounds fine thanks. I initially thought you were talking about another private subnet to allow internal Mifos to ML but the pub IP with port 80 and 443 open for inbound will be great! Can I assume you will also assign an elastic Pub IP to the Mifos VM(s) ?