This documents how to use Ansible playbooks to set up a production-like server installation. It differs from a production installation in that certificates must not be self-signed in a production environment.
These steps are for installing on a server OS directly and require experience with remote configuration and Linux administration.
You must have a local VM or remote server. See
/packaging/vagrant/centos for a Vagrant VM (CentOS 7) script for working example of creating a local VM.
Clone the main repository. The Ansible playbooks and templates are in that folder.
git clone https://github.com/intrahealth/client-registry.git
Create a VM. Make sure to include a public ssh key for the user who will install prerequisites. Your SSH public key should be in
.ssh/authorized_keys on the remote host, ie:
cat ~/.ssh/id_rsa.pub | ssh user@remotehost 'cat >> .ssh/authorized_keys'
Hosts can be specified in inventory files or on the command line. To use Ansible with an inventory file, you must create a file or edit the one in the repository. There are yaml and ini formats supported.
hosts file that has an entry for one server would be:
[servers]is not necessary, it is way to tag groups of servers. The file may simply contain an IP address or domain.
To use the hosts file:
ansible-playbook -i hosts someplaybook.yaml
Alternately, hosts may be specified on the command line (the comma is necessary even if there is only one host):
ansible-playbook -i 172.16.168.158, someplaybook.yaml
opencr user (optional)
A example playbook is provided to show how to create a
opencr user with sudo permissions using Ansible to be used with the host.
opencr user and gives it sudo access:
ansible-playbook -i hosts user.yaml
ansible-playbook -i hosts prep_centos.yaml -e user=opencr
ansible-playbook -i hosts elasticsearch.yaml -e user=opencr
ansible-playbook -i hosts tomcat.yaml -e user=opencr
ansible-playbook -i hosts postgres.yaml -e user=opencr -e pgpass=hapi
ansible-playbook -i hosts hapi.yaml -e user=opencr
ansible-playbook -i hosts opencr.yaml -e user=opencr
An optional step but recommeded is to check the logs for services running after installation:
ansible-playbook -i hosts troubleshoot.yaml -e user=opencr
OpenCR is now running. It will only allow requests from localhost (from the same server it is installed on).
HTTPS must be used.
If not running localhost, follow the next steps to create self-signed server and client certs, and copy them onto the server using an Ansible script below.
Certificates (Required if not using localhost)
These steps are automated in the certs.sh script, but please read through the steps to understand what is happening.
bash certs.sh <ip address or domain>
Then run the ansible script to replace it on the server:
ansible-playbook -i hosts servercerts.yaml -e user=opencr
Two certificate pairs are required, one pair for the server and one for the client generated from the server's. The existing self-signed server certs use localhost as the CN. This can be seen with the following for any cert:
$ openssl x509 -in ../../server/certificates/server_cert.pem -text
Subject: CN = localhost, O = Client Registry
This means that new server and client certificates need to be generated with the IP address or domain for clients to access the client registry if it is not running on localhost.
Self-signed certificates must only be created for testing and demonstrations and in non-production settings.
Make a note of your IP or domain for which you need to create a server cert. Run the following to create a new server cert/key pair. It will ask for a pass phrase (which is required in production) but -nodes option squashes that. This will create two files, server_key.pem and server_cert.pem. We can inspect the certificate to verify it has the IP address in the subject.
# confirm ip being used
openssl req -x509 -newkey rsa:4096 -keyout server_key.pem -out server_cert.pem -days 365 -subj "/CN=172.16.168.172" -nodes
# confirm new CN
openssl x509 -in server_cert.pem -text
Now it is necessary to create new a new client cert based on the server cert. A key is first created, then the certificate, and they are packaged together in a p12 file.
openssl req -newkey rsa:4096 -keyout ansible_key.pem -out ansible_csr.pem -nodes -subj "/CN=ansible"
openssl x509 -req -in ansible_csr.pem -CA server_cert.pem -CAkey server_key.pem -out ansible_cert.pem -set_serial 01 -days 36500
# requires specifying an export key
openssl pkcs12 -export -in ansible_cert.pem -inkey ansible_key.pem -out ansible.p12
The client certs can be placed in the existing folder for client certs for convenience.
# add client certs
cp ansible_key.pem ../../server/sampleclientcertificates/
cp ansible_csr.pem ../../server/sampleclientcertificates/
cp ansible_cert.pem ../../server/sampleclientcertificates/
cp ansible.p12 ../../server/sampleclientcertificates/
Server certs may also be copied into ../../certificates/ for convenience but this will overwrite the copies and break localhost if the repo is used to upload code to GitHub.
To complete the process, server certs need to be placed into the server. This will be done using an Ansible script below.
Copy the server certs to the server and restart opencr service to use them.
ansible-playbook -i hosts servercerts.yaml -e user=opencr
Test (for servers not localhost):
# this assumes the server cert and client cert are in this (/packaging/ansible) directory
# replace the path to your copy of the repo
# replace the ip address of the server
curl --cert ansible.p12 --cert-type p12 --cacert server_cert.pem -d @/Users/richard/src/github.com/intrahealth/client-registry/DemoData/patient1_openmrs.json -H "Content-Type: application/json" -XPOST https://172.16.168.172:3000/Patient
Add additional user public keys
As necessary, add additional ssh keys to the user
opencr. (Ensure that the user's public key is available on github, ie. https://github.com/citizenrich.keys):
ansible-playbook -i hosts keys.yaml