Eduroam tools container-based deployment: Overall considerations

The ancilliary tools package consists of three separate sets of tools:

admintool
metrics
monitoring

Each of the tools is (at the moment) designed to run in an isolated environment. They can be run on a single docker host by mapping each to a different port. The configuration files provided here are designed this way:

Admintool runs on ports 80 and 443 (HTTP and HTTPS)
Monitoring tools run on ports 8080 and 8443 (HTTP and HTTPS)
Metrics runs on ports 9080 and 9443 (HTTP and HTTPS)

It is possible (and recommended if the resources are available) to run these on separate hosts - in which case each can run on the standard port HTTP/HTTPS ports 80/443. Details on the configuration change needed are given below.

ChangeLog

Changes to this document since the workshop at APAN41 in Manilla, January 2016.

2020-03-10: Add NRO_ROID, update images for EDB2.
2020-03-10: Add Places API to Google Maps instructions.
2019-02-15: Remove google-plus from supported login methods.
2019-02-07: Google Login now longer needs Google+ API (being removed).
2019-02-07: Icinga uses SITE_PUBLIC_HOSTNAME also for outgoing notifications.
2018-08-09: Added documentation on generating radsecproxy.conf
2018-08-09: Update Google Maps API key documentation (billing now required)
2018-08-09: Add documentation for sending logs from radsecproxy to ELK.
2018-08-09: Troubleshooting: list volumes, new name for elasticsearch data volume.
2018-08-03: Misc doc updates: Icinga conf from admin tool, ELK setup, VM settings for ELK, Google credentials.
2018-07-30: Added instructions to create user profile for service accounts in Admintool.
2018-03-02: Added environment variable FEDERATION_DOC_URL.
2017-03-17: Added environment variables for institutional branding.
2016-08-30: Added instructions for configuring a GoogleMaps API key.
2016-06-14: Updated Admintool documentation: added instructions for generating a CSR.
2016-06-14: Updated Metrics documentation: replace LOCAL_INST env var with INST_NAMES, describe visualizations.
2016-06-10: Updated documentation on metrics service: installing and accessing visualizations.
2016-06-02: Updated documentation on metrics service: new variable LOCAL_COUNTRY.
2016-05-05: Added documentation on feeding radius logs into ELK.
2016-04-08: Added documentation on entering server details for monitoring into the Admintool.
2016-04-06: Added system recommendations.
2016-04-05: Added documentation on changing settings after running setup scripts.
2016-04-05: Added Troubleshooting section.
2016-04-05: Make REALM_EXISTING_DATA_URL default to blank.
2016-04-05: Added documentation on creating local accounts.
2016-04-05: Added section documenting Network access requirements.
2016-04-05: Added documentation for adjusting port numbers.
2016-04-05: Added documentation for installing proper SSL certificates.
2016-04-05: Added Metrics tools authentication configuration.
2016-04-04: Added Admintool settings for NREN Institution and Federation name.
2016-04-01: Added instructions for loading Icinga configuration generated by Admintool.
2016-04-01: Added Update instructions.
2016-04-01: Added this ChangeLog section.
2016-03-08: Added documentation for passing arbitrary configuration options to Admintool.
2016-03-08: Added configurable login methods for Admintool.
2016-02-25: Added ADMINTOOL_SECRET_KEY environment variable.
2016-02-15: Added ADMINTOOL_DEBUG environment variable.
2016-02-15: Clarified deployment instructions.
2016-02-15: Added missing instructions to enable Google+API.

Preliminaries - System considerations

All of the services are packaged as Docker containers, so are not tightly linked to the details of the system they are running on. However, the following considerations apply to the system:

The system should be running a recent version of Linux supported by Docker. For Ubuntu, this is 15.10 / 14.04 / 12.04.
The system can be either a virtual machine or a physical (real hardware) system.
The system should have at least 5GB of disk space (recommended 10GB) available under /var/lib/docker (on the /var partition or on the root filesystem if not partitioned).
The system should have at least 4GB of RAM (recommended 8GB) available.

Preliminaries - Docker

Install and configure Docker. Please follow our Docker setup instructions.

Please become familer with Docker by following our Docker introduction.

Preliminaries - Mail server

Some of the tools (admintool and monitoring) will need to send outgoing email. Please make sure you have the details of an SMTP ready - either one provided by your systems administrator, or one running on the local system.

Eduroam ancillary tools: Basic setup

NOTE: For the XeAP workshop at APAN41, please follow instructions in Lab.md

The installation instructions here are ment for deployment at your institution - the ones in Lab.md are targeted for the lab VMs.

On each of the VMs, start by cloning the git repository:

git clone https://github.com/REANNZ/etcbd-public

Deploying admintool

Modify the admintool.env file with deployment parameters - override at least the following values:

Pick your own admin password: ADMIN_PASSWORD
Pick internal DB passwords: DB_PASSWORD, POSTGRES_PASSWORD
- Generate these with: openssl rand -base64 12
SITE_PUBLIC_HOSTNAME: the hostname this site will be visible as.
LOGSTASH_HOST: the hostname the metrics tools will be visible as.
EMAIL_* settings to match the local environment (server name, port, TLS and authentication settings)
SERVER_EMAIL: From: email address to use in outgoing notifications.
ADMIN_EMAIL: where to send NRO admin notifications.
REALM_COUNTRY_CODE / REALM_COUNTRY_NAME - your eduroam country
NRO_ROID: Roaming Operator ID, as assigned by eduroam-OT. Defaults to <REALM_COUNTRY_CODE>01.
Note that this option is not included in the file template, add it manually if there is a need to override the default value.
NRO_INST_NAME: the name of the institution acting as the National Roaming Operator (NRO)
NRO_INST_URL: the URL of the institution acting as the NRO
NRO_INST_FACEBOOK (OPTIONAL): the Social Media (Facebook) handle of the NRO institution
NRO_INST_TWITER (OPTIONAL): the Social Media (Twitter) handle of the NRO institution
NRO_FEDERATION_NAME: the name of the AAI federation the Admintool is connected to (if available to the NRO). Leave unmodified if no AAI federation exists locally.
FEDERATION_DOC_URL: URL to the federation policy document / documentation about the federation.
TIME_ZONE - your local timezone (for the DjNRO web application)
MAP_CENTER_LAT, MAP_CENTER_LONG - pick your location
REALM_EXISTING_DATA_URL: In a real deployment, leave this set to blank (REALM_EXISTING_DATA_URL=) to start with an empty database. In a test deployment, you can point to an existing database to load the initial data (e.g., you can load the REANNZ data by setting the URL to https://member.eduroam.net.nz/general/institution.xml).
GOOGLE_KEY/GOOGLE_SECRET - provide Key + corresponding secret for an API credential (see below on configuring these settings)
GOOGLE_API_KEY - provided a Google Maps browser API key (see below)
ADMINTOOL_SECRET_KEY: this should be a long and random string used internall by the admintool.
- Please generate this value with: openssl rand -base64 48
ADMINTOOL_DEBUG: for troubleshooting, uncomment the existing line: ADMINTOOL_DEBUG=True - but remember to comment it out again (or change to False or blank) after done with the troubleshooting.
ADMINTOOL_LOGIN_METHODS: enter a space-separated list of login methods to enable.
- Choose from:
  - shibboleth: SAML Login with Shibboleth SP via an identity federation. Not supported yet.
  - locallogin: Local accounts on the admin tool instance.
    - Note that local accounts can later be created by logging into the Admintool at https://admin.example.org/admin/ as the administrator (with the username and password created here), and selecting Users from the list of tables to administer, and creating the user with the Add user button.
  - google-oauth2: Login with a Google account. Only works for applications registered with Google - see below on enabling Google login.
  - yahoo: Login with a Yahoo account. No registration needed.
  - amazon: Login with an Amazon account. Registration needed.
  - docker: Login with a Docker account. Registration needed.
  - dropbox-oauth2: Login with a Dropbox account. Registration needed.
  - facebook: Login with a Facebook account. Registration needed.
  - launchpad: Login with an UbuntuOne/Launchpad account. No registration needed.
  - linkedin-oatuh2: Login with a LinkedIn account. Registration needed.
  - meetup: Login with a MeetUp account. Registration needed.
  - twitter: Login with a Twitter account. Registration needed.
- Please note that many of these login methods require registration with the target site, and also need configuring the API key and secret received as part of the registration. Please see the Python Social Auth Backends documentation for the exact settings required.
Additional settings: it is also possible to pass any arbitrary settings for the Admintool (or its underlying module) by prefixing the setting with ADMINTOOL_EXTRA_SETTINGS_. This would be relevant for passing configuration entries to the login methods configured above - especially SECRET and KEY settings for login methods requiring these. Example: to pass settings required by Twitter, add them as:
```
  ADMINTOOL_EXTRA_SETTINGS_SOCIAL_AUTH_TWITTER_KEY=93randomClientId
  ADMINTOOL_EXTRA_SETTINGS_SOCIAL_AUTH_TWITTER_SECRET=ev3nM0r3S3cr3tK3y
```
Icinga configuration: please see the separate section on using the Admintool to generate configuration for Icinga.

Note this file (admintool.env) is used both by containers to populate runtime configuration and by a one-off script to populate the database.

As an additional step, in global-env.env, customize system-level TZ and LANG as desired.

Use Docker-compose to start the containers:

cd etcbd-public/admintool
docker-compose up -d

Run the setup script:

./admintool-setup.sh admintool.env

Please note: the admintool-setup.sh script should be run only once. Repeated runs of the script would lead to unpredictable results (some database structures populated multiple times). Also, for most of the configuration variables (except those listed below), re-running the script is not necessary - restarting the containers should be sufficient (docker-compose up -d). The following variables would are the ones where a container restart would NOT be sufficient:

SITE_PUBLIC_HOSTNAME: if this value has changed after running admintool-setup.sh, besides restarting the containers, change the value also in the Sites entry in the management interface at https://admin.example.org/admin/
REALM_COUNTRY_CODE, NRO_INST_NAME: if any of these values has changed after running admintool-setup.sh, besides restarting the containers, change the value also in the Realms entry in the management interface at https://admin.example.org/admin/
ADMIN_USERNAME, ADMIN_EMAIL, ADMIN_PASSWORD: if any of these values has changed after running admintool-setup.sh, besides restarting the containers, change the value also in the account definition in the Users table in the management interface at https://admin.example.org/admin/
DB_USER, DB_NAME, DB_PASSWORD, DB_HOST: changing these values after database initialization is an advanced topic beyond the scope of this document. Please see the Troubleshooting section below.
REALM_EXISTING_DATA_URL: changing this value after database initialization is an advanced topic beyond the scope of this document. Please see the Troubleshooting section below.

At this point, please become familiar with Docker-compose by following our Introduction to Docker-compose:

Installing proper SSL certificates for Admintool

We strongly recommend operating the Admintool with proper SSL certificates. The tool would generate a self-signed certificate on startup in order to operate, but this would make interaction difficult for both end-users (browser access) as well as for automated tools.

We leave obtaining the certificates outside the scope of this document.

Obtain the certificates in PEM format, following your usual procedure for obtaining certificates.
- In the absence of usual procedure, generate a certificate signing request with:
```
  openssl req -new -newkey rsa:2048 -sha256 -keyout server.key -nodes -out server.csr
```
- Submit the CSR to your preferred Certification Authority and wait to receive a server certificate as server.crt.
Combine the server certificate with any intermediate certificates (if provided by the CA) by concatenating the server certificate with any intermediate certificates into a single file (server certificate first, then any intermediate certificates ordered leaf to root). Store this as server.crt in a directory on the server running the Admintool.
Store the private key as server.key in the same directory.

Copy the two files into the Apache certificates volume with:

  docker run -v `pwd`:/certs-in -v admintool_apache-certs:/apache-certs -it --rm --name debcp debian:jessie cp /certs-in/server.crt /certs-in/server.key /apache-certs

Preparing Icinga configuration in Admintool

In the next section, we will be deploying the monitoring tools (Icinga). The monitoring tools will be loading the configuration generated by the Admintool. We need to configure this before we proceed to installing the monitoring tools.

The generated configuration will:

Define an Icinga Host for each NRO Radius server.
Define an Icinga Host for each Institutional Radius server.
Configure a Ping connectivity check for each server.
If supported by the Radius server, generate a Radius Status Check for the server.
For each Monitored Realm with authentication configured, there will be:
- A service check through each NRO server.
- A service check through each Institutional Radius server that is listed as a proxy for the Institutional Realm.
Each service check will be configured with Notifications to Contacts associated with Institution the Institutional Realm being checked belongs to.
For each Institutional Radius server, the Notifications would also go to all Contacts associated with all Institutions the server is associated with.

There is a number of configuration variables in admintool.env that control how the configuration will be generated:

NRO_SERVERS specifies the list of NRO servers (short identifiers) - example: NRO_SERVERS=server1 server2
For each server, there should be a set of variables using the short server identifier as part of the name:
- NRO_SERVER_HOSTNAME_serverid: the hostname (or IP address) to use for server checks
- NRO_SERVER_SECRET_servererid: the Radius secret to use with the server
- NRO_SERVER_PORT_serverid: the Radius port number to use with this server (Optional, defaults to 1812)
- NRO_SERVER_STATUS_serverid: should be set to True if the server supports Radius Status checks (Optional, defaults to False)
ICINGA_CONF_REQUEST_CUI: should be set to True if the eduroam login checks should request the Chargeable User Identity (CUI) attribute (Optional, defaults to True)
ICINGA_CONF_OPERATOR_NAME: the Operator Name to use in the eduroam login checks (Optional, no operator name is passed if not specified).
ICINGA_CONF_VERBOSITY: the verbosity level in eduroam login checks. (Optional, defaults to 1. Level of at least 1 is needed to see the CUI returned by the check. Values can range from 0 to 2).
ICINGA_CONF_GENERATE_INSTSERVER_CHECKS: Should the generated configuration include the institutional server checks? (Optional, defaults to False).
ICINGA_CONF_NOTIFY_INST_CONTACTS: Should the generated configuration notify institutional contacts - for server and monitored realm checks? (Optional, defaults to True). This setting gives the option to disable all Notifications sent to institional contacts if not desired - in which case alerts would go only to the nominated NRO email address given in the Monitoring tools configuration.
Update Admintool to pick up the changed configuration:
```
  docker-compose up -d
```

As a final step, prepare the account Icinga would use to retrieve the configuration:

Log into the Admintool admin interface at https://admin.example.org/admin/ as the administrator (with the username and password created earlier).
Select Users from the list of tables to administer.
Use the Add user button to bring up the user creation form.
Enter the username and password
- The sample Monitoring configuration uses confuser as the username
- We recommend generating the password, e.g. with `openssl rand -base64 12``
Use the Save and continue editing button to get to the next screen with additional details
- In the list of Available user permissions, select all three edumanage | Monitored Realm (local authn) permissions and add them to the Chosen user permissions. (The permission to access the monitoring credentials is internally used to represent as permission to access the monitoring configuration).
- Select Save to store the permissions.
We also recommend to create a User Profile entry for this user account - to keep the internal database consistent with assumptions the DjNRO code makes:
- Select User Profiles from the list of tables to administer (navigate back to the Home screen to get the list of tables).
- Use the Add User Profile button to bring up the user profile creation form.
- Select the user just created in the above step as the user.
- Select an institution (the NRO organisation would do) to associate the user with.
- Leave the "Is social active" option unchecked.
  - This way, the account does NOT get permission to administer the institution - but a user profile is created, making the internal database consistent.
- Save the user profile.

Now you should be able to access the monitoring configuration at https://admin.example.org/icingaconf with this account. (And it is also accessible under the administrator account).

Entering data for monitoring configuration into the Admintool.

The following information needs to be entered into the Admintool in order to generate the monitoring configuration. Due to dependencies between different data objects, we recommend entering the data in the order given here.

All of the data can be entered at https://admin.example.org/admin/

Add an Institution - select the NRO Realm, entity type and enter the English name of the Institution.
Add at least one Contact for the institution (not that Contacts can be reused across institutions).
Add an Institution Details object: select the Institution it belongs to, enter Address details and select at least one Contact.
Add an Institution Server: select the institution (or possibly more) the server belongs to, server type, a descriptive name and the host name, the Radius secret, and select whether the server responds to Status messages.
Add the Institution's Realms: enter the Realm name, select the Institution and select the institution's radius server that the NRO radius server should be proxying to.
Add an Institution Monitored Realm entry: select the Institution's Realm and select "Local account authentication" as the Monitor type (the only available entry).
Add a Monitored Realm (local authn) entry: select the Monitored Realm and select the EAP method, phase2 authentication, username and password to match the institution's radius server settings.

Entering all of the above should be sufficient for the Admintool to generate monitoring checks for Icinga - both via the NRO servers and directly through the Institution's servers.

Deploying monitoring tools

Modify the icinga.env file with deployment parameters - override at least the following values:

Pick your own admin password: ICINGAWEB2_ADMIN_PASSWORD
Pick internal DB passwords: ICINGA_DB_PASSWORD, ICINGAWEB2_DB_PASSWORD, POSTGRES_PASSWORD
- Generate these with: openssl rand -base64 12
SITE_PUBLIC_HOSTNAME: the hostname this site will be visible as (also used by Icinga for outgoing notifications)
ICINGA_ADMIN_EMAIL: where to send notifications
EMAIL_* settings to match the local environment (server name, port, TLS and authentication settings)
EMAIL_FROM - From: address to use in outgoing notification emails.
The following configure Icinga monitoring and check and notification intervals:
- ICINGA_NOTIFICATION_INTERVAL: Re-notification interval in seconds. 0 disables re-notification. Defaults to 7200 (2 hours).
- ICINGA_CHECK_INTERVAL: Check interval in seconds (HARD state). Defaults to 300 (5 minutes)
- ICINGA_RECHECK_INTERVAL: Interval (in seconds) for rechecking after a state change (SOFT state). Defaults to 60 seconds.
The following settings configure how Icinga fetches the configuration generated by the Admintool:
CONF_URL_LIST: the URL (possibly a list of URLs) to fetch configuration from. Should be https://eduroam-admin.example.org/icingaconf
CONF_URL_USER: the username to use to authenticate to the configuration URL.
CONF_URL_PASSWORD: the password to use to authenticate to the configuration URL.
CONF_REFRESH_INTERVAL: the time period (in seconds) to wait before reloading the configuration from the URL. Defaults to 3600 (1 hour).
WGET_EXTRA_OPTS: additional options to passs to wget when fetcing the configuration. This needs in particular to take care of wget establishing trust for the certificate presented by the server.
- If the Admintool Apache server has already been configured with a certificate from an accredited Certification Authority (optional step above), no further action is needed.
- If the Admintool is using a self-signed automatically generated certificate, we recommend:
  - Copying this certificate into the Icinga configuration volume so that wget can access the certificate:
    - Note that at this stage, the volume for the Icinga external configuration has not been created yet, so we need to jump the gun and create it explicitly (otherwise, the volumes get created automatically by docker-compose up): docker volume create --name icinga_icinga-external-conf
    - Copy the certficiate (by running the "cp" command in a container mounting both volumes): docker run --rm --name debian-cp -v admintool_apache-certs:/admintool-apache-certs -v icinga_icinga-external-conf:/icinga_icinga-external-conf debian:jessie cp /admintool-apache-certs/server.crt /icinga_icinga-external-conf/
  - And instructing wget to trust this certificate: WGET_EXTRA_OPTS=--ca-certificate=/etc/icinga2/externalconf/server.crt
  - Alternatively, it is also possibly to instruct wget to blindly accept any certificate - but as this creates serious security risks, it can only be used for internal tests and MUST NOT be used in production. The setting is: WGET_EXTRA_OPTS=--no-check-certificate

This file is used by both the containers to populate runtime configuration and by a one-off script to populate the database.

Additionally, in global-env.env, customize system-level TZ and LANG as preferred - or you can copy over global.env from admintool:

cd etcbd-public/icinga
cp ../admintoool/global-env.env .

Use Docker-compose to start the containers:

cd etcbd-public/icinga
docker-compose up -d

Run the setup script:

./icinga-setup.sh icinga.env

Please note: the icinga-setup.sh script should be run only once. Repeated runs of the script would lead to unpredictable results (some database structures populated multiple times). Also, for most of the configuration variables (except those listed below), re-running the script is not necessary - restarting the containers should be sufficient (docker-compose up -d). The following variables would are the ones where a container restart would NOT be sufficient:

ICINGAWEB2_ADMIN_USERNAME, ICINGAWEB2_ADMIN_PASSWORD: if any of these values has changed after running icinga-setup.sh, change the value also in the account definition in the management interface at https://monitoring.example.org/icingaweb2/ (navigate to Configuration => Authentication => Users => select the account).
ICINGA_DB_USER, ICINGA_DB_NAME, ICINGA_DB_PASSWORD, ICINGA_DB_HOST, ICINGAWEB2_DB_USER, ICINGAWEB2_DB_NAME, ICINGAWEB2_DB_PASSWORD, ICINGAWEB2_DB_HOST: changing these values after database initialization is an advanced topic beyond the scope of this document. Please see the Troubleshooting section below.

Note: at this point, Icinga will be executing checks against all Radius servers as configured. It is essential to also configure the Radius servers to accept the Icinga host as a Radius client - with the secret as configured in the Admintool.

Note: Icinga will be using the configuration as generated by the Admintool. When the settings change in Admintool, Icinga would only see the changes next time it fetches the configuration. This defaults to happen every hour. To trigger an immediate reload, either restart the container, or send it the "Hang-up" (HUP) signal:

docker kill --signal HUP icinga

Installing proper SSL certificates for Monitoring tools

We also recommend operating the Monitoring tools with proper SSL certificates.

We leave obtaining the certificates outside the scope of this document.

Obtain the certificates in PEM format (see the above section on certificates for Admintool for more detailed instructins)..
Combine the server certificate with any intermediate certificates (concatenated into a single file) and store this as server.crt in a directory on the server running the Monitoring tools (again, see the above section on certificates for Admintool for more detailed instructins).
Store the private key as server.key in the same directory.

Copy the two files into the Apache certificates volume with:

  docker run -v `pwd`:/certs-in -v icinga_icingaweb-apache-certs:/apache-certs -it --rm --name debcp debian:jessie cp /certs-in/server.crt /certs-in/server.key /apache-certs

Deploying metrics tools

Modify the elk.env file with deployment parameters - override at least the following values:

Pick your own admin password: ADMIN_PASSWORD
ADMIN_USERNAME: optionally configure the administrator username (default: admin)
SITE_PUBLIC_HOSTNAME: the hostname this site will be visible as.
ADMIN_EMAIL: email address so far only used in web server error messages.
LOCAL_COUNTRY: two-letter country code of the local country (for metrics to identify domestic and international visitors and remote sites).
INST_NAMES: white-space delimited list of domain names of institutions to generate per-instititon dashboard and visualizations for.

Additionally, in global-env.env, customize system-level TZ and LANG as preferred - or you can copy over global.env from admintool:

cd etcbd-public/elk
cp ../admintoool/global-env.env .

ElasicSearch needs one system-level setting tuned: vm.max_map_count needs to be increased from the default 65530 to 262144. Without this setting in place, ElasticSearch fails to start. To change the setting both in the currently booted system and to make it apply after a reboot, run (as root):

sysctl -w vm.max_map_count=262144
echo "vm.max_map_count = 262144" > /etc/sysctl.d/70-elasticsearch.conf

And now, back in etcbd-public/elk, use Docker-compose to start the containers:

docker-compose up -d

Run the setup script:

./elk-setup.sh

Please note: the elk-setup.sh script may be run multiple times (contrary to the other setup scripts). This script pushes a small set of predefined visualizations and dashboards into the Metrics tool (the Kibana web application) - and also applies some setting to elasticsearch (disable replicas, tune down logging). The only environment variables this script depends on are LOCAL_COUNTRY and INST_NAMES: to propagage a change to these variable, it is necessary to first restart the containers (docker-compose up -d), then re-run the setup script. (If institutions were removed from INST_NAMES, it may be necessary to manually remove the dashboard and visualizations created for these institutions).

The setup script also creates an index pattern in Kibana, telling Kibana where to find your data. However, this may fail if there are no data in ElasicSearch yet, so you may have to reinitialize Kibana after some initial data is ingested into ElasticSearch. You can re-run the setup script with:

./elk-setup.sh --force

Note that the --force flag deletes all Kibana settings - but the initial ones get loaded again by the setup script. And the actual data in ElasicSearch stays intact.

Installing proper SSL certificates for Metrics tools

We also recommend operating the Metrics tools with proper SSL certificates.

We leave obtaining the certificates outside the scope of this document.

Obtain the certificates in PEM format (see the above section on certificates for Admintool for more detailed instructins).
Combine the server certificate with any intermediate certificates (concatenated into a single file) and store this as server.crt in a directory on the server running the Metrics tools (again, see the above section on certificates for Admintool for more detailed instructins).
Store the private key as server.key in the same directory.

Copy the two files into the Apache certificates volume with:

  docker run -v `pwd`:/certs-in -v elk_apache-certs:/apache-certs -it --rm --name debcp debian:jessie cp /certs-in/server.crt /certs-in/server.key /apache-certs

Feeding Radius logs from NRS servers (radsecproxy) into metrics tools (ELK)

The main benefits in having the metrics tools is in using them to explore the usage data from the eduroam Radius servers.

The right collection point is at the NRS servers.

Setting up the NRS servers is outside the scope of this documentation - the prerequisites are that:

NRS servers are already deployed with radsecproxy (either manual install or via Docker). If you are running freeradius instead, pleasee see the next section instead.
Radsecproxy has been built with the patch to log OperatorName and CUI.
The radsecproxy log file location is /var/log/radsecproxy/radsecproxy.log

We recommend to install filebeat to ship the logs to the ELK stack. We recommend using Docker to deploy filebeat. Part of this repository are also files instructing Docker to run filebeat.

The steps are:

Install Docker CE on the radius (NRS) server. Please follow our Docker setup instructions.
Check out this repository (same as when deploying the ancillary service as per above) - and go to the filebeat-radsecproxy directory:
```
  git clone https://github.com/REANNZ/etcbd-public.git
  cd etcbd-public/filebeat-radsecproxy/
```
Modify the filebeat-radsecproxy.env file with deployment parameters:
- LOGSTASH_HOST: the hostname the metrics tools are reachable at.
- RADIUS_SERVER_HOSTNAME: the hostname of the radius server. (This will be used in the metadata of the log messages).
And in global-env.env:
- Set TZ to the same timezone as your radsecproxy. This is important for getting timestamps processed correctly.
- And set LANG to your preferred locale (or you can leave this one as is).
Use Docker-compose to start the containers - and watch the logs to confirm it works as it should:
```
  cd etcbd-public/filebeat-radsecproxy
  docker-compose up -d
  docker-compose up logs -f
```
Check metrics now see the usage data from your server: https://xx-rad1.tein.aarnet.edu.au:9443/
- Explore raw data ("Discover")
- Visualizations
- Dashboards
- NOTE: you may have to make Kibana rescan the available after ingesting first set of data. While this can be also done inside Kibana, you can also re-run the setup script (on the host where the metrics tools are deployed) with:
```
  ./elk-setup.sh --force
```

Feeding Radius logs from NRS servers (freeradius) into metrics tools (ELK)

As of August 2018, you should be deploying your NRS servers with radsecproxy - and use the radsecproxy log forwarder documented in the previous section to ship logs to the metrics tools.

In case you are running freeradius on your NRS server, this section provides the freeradius-specific documentation for shipping the logs to the metrics tools. If you are not running freeradius on your NRS server, please skip this section.

This section assumes that:

NRS servers are already deployed.
NRS servers are configured to produce the linelog log file.
The linelog file includes the Chargeable User Identity (CUI) field.
The linelog file location is /var/log/freeradius/linelog

We then recommend to install filebeat to ship the logs to the ELK stack. We recommend using Docker to deploy filebeat. Part of this repository are also files instructing Docker to run filebeat.

The steps are:

Install Docker Engine on the radius (NRS) server. Please follow our Docker setup instructions.
Check out this repository (same as when deploying the ancillary service as per above):
```
  git clone https://github.com/REANNZ/etcbd-public
```
Modify the filebeat-radius.env file with the following deployment parameters:
- LOGSTASH_HOST: the hostname the metrics tools are reachable at.
- RADIUS_SERVER_HOSTNAME: the hostname of the radius server (how it should be reported to the metrics tools)
And in global-env.env:
- Set TZ to the same timezone as your radsecproxy. This is important for getting timestamps processed correctly.
- And set LANG to your preferred locale (or you can leave this one as is).
Use Docker-compose to start the containers - and watch the logs to confirm it works as it should:
```
  cd etcbd-public/filebeat-radius
  docker-compose up -d
  docker-compose logs -f
```

Accessing visualizations in metrics tools

After properly setting up the Metrics tool as per above, it should come with a small initial set of visualizations and three dashboards.

These can be found under the Dashboards menu (in the main navigation side-bar) - or alterntively, can also be accessed directly at

In either case, it will be necessary to select an appropriate time range through the Kibana time selector tool (top right corner of the browser window).

The visualizations installed (and used in the above dashboards) are:

IdP-Trend-Aggregate: graph of daily authentication sessions, split into domestic and international based on the IdP the user authenticated through.
SP-Trend-Aggregate: graph of daily authentication sessions, split into domestic and international based on the SP the user authenticated to.
VisitorsInclDomestic: per-country composition of authenticated session based on country of origin (IdP the user authenticated through)
VisitorsExclDomestic: per-country composition of authenticated session based on country of origin (IdP the user authenticated through), excluding domestic users (where realm_country equals $LOCAL_COUNTRY)
IdP-Trend-Institution_<inst-dns-name>: per-institution graph of daily authentication sessions where this institution was the IdP, split into domestic and international based on the site the user authenticated to.
SP-Trend-Institution_<inst-dns-name>: per-institution graph of daily authentication sessions where this institution was the SP, split into domestic and international based on the IdP the user authenticated through.
IdP-Activity-Institution_<inst-dns-name>: per-institution graph of daily authentication sessions where this institution was the IdP, split per domain (operator_name) of the site the user authenticated to.
SP-Activity-Institution_<inst-dns-name>: per-institution graph of daily authentication sessions where this institution was the SP, split per domain (realm) of the IdP the user authenticated through.

Troubleshooting visualizations

If you get an error message about variables not being found when displaying the above visualizations, this is likely because the Kibana Index Pattern was created before all of the fields were introduced into Kibana (through having log messages arrive via logstash).

The very simple solution is to re-run the elk-setup.sh script with the --force flag - telling it to delete all Kibana settings (stored in the .kibana index in elasicsearch) - but the initial Kibana settings get loaded again by the setup script. And the actual data in ElasicSearch stays intact.

cd etcbc-public/elk
./elk-setup.sh --force

Alternative to that is to re-create the Index Pattern from within Kibana: in Kibana, navigate to Management -> Index Patterns and under the existing logstash-* index pattern, select the "Refresh" icon (and confirm the action). This would rescan the indices available and would introduce any newly found variables into Kibana.

Accessing the services

The Admintool can be accessed at https://admin.example.org/

The management interface of the Admnintool can be accessed at https://admin.example.org/admin/

The management interface of the monitoring tools (Icingaweb2) can be accessed at https://monitoring.example.org:8443/icingaweb2/

The web interface of the metrics tools (Kibana) can be accessed at https://metrics.example.org:9443/

Configuring port numbers

The configuration for the three services has been done to allow them to run on the same host (avoiding port clashes) - but that means Monitoring and Metrics web services run on nonstandard port numbers.

If the tools are deployed on separate hosts, it would be desirable to run them on standard port numbers (as there is no longer the need to avoid port clashes).

To make the Monitoring tools run on standard port numbers: edit icinga/docker-compose.yml and change the port mappings for the icingaweb service from:

ports:
  # temporarily use alt ports to avoid clash with admintool
  - "8080:80"
  - "8443:443"

to:

ports:
  - "80:80"
  - "443:443"

and also change the HTTPS_PORT variable to match the new port number - i.e., from:

environment:
    HTTPS_PORT: 8443

to:

environment:
    HTTPS_PORT: 443

Restart the containers with:

cd etcbc-public/monitoring
docker-compose up -d

To make the Metrics tools run on standard port numbers, edit elk/docker-compose.yml and change the port mappings for the apache service and the HTTPS_PORT environment variable the same way (and also restart the containers.

Network access requirements

The Admintool, Monitoring, and Metrics tools provide a variety of services with different audiences - the following lists access that should be granted (both at the perimeter firewall and in the local host firewall as applicable):

The Admintool is to be accessed by NRO staff, institutional administrators, and the general public - so should be publicly accessible on TCP ports 80 and 443.
The Monitoring and Metrics tools would be accessed primarily by NRO staff over HTTP/HTTPS. Access is protected by username/password authentication over HTTPS - so depending on local policies and preferences, the HTTP+HTTPS ports can be made accessible either publicly, or just from the NRO internal network.
The Metrics tools also listen on TCP port 5043 for incoming messages from other services (primarily the Radius server, also Apache on the Admintool). This port should be made accessible only to the services pushing in data and MUST NOT be publicly accessible (i.e., MUST be protected by a firewall).

Updating deployed tools

Over time, these tools will receive updates - as upstream software releases new version, as new features are developed, or as bugs and security issues are fixed.

The updates would target both the container images and the files driving the images contained in these repositories.

It is essential to install these updates to continue operating these tools in a reliable and secure manner.

Updating Docker files

The first step is to pull updates to the Docker (and docker-compose) files driving the tools - i.e., this repository:

cd etcbd-public
git fetch --verbose --all
git merge origin/master

If the above commands succeed, this part is completed. Git may fail with an error message like:

error: Your local changes to the following files would be overwritten by merge:
        admintool/admintool.env
        Please, commit your changes or stash them before you can merge.
        Aborting

In that case:

Save your changes to the file(s) affected into the git "stash" with: git stash save
Now merge the updates to the original (unmodified) files: git merge origin/master
Now bring back the changes from the stash: git stash pop
Git tries to merge the local modifications into the updated files.
If this succeeds without encountering any conflicts, this step is done.

If git reports conflicts - like:

  $ git stash pop
  Auto-merging admintool/admintool.env
  CONFLICT (content): Merge conflict in admintool/admintool.env

It is necessary to manually resolve the conflict. Edit the file and manually merge the updates (typically adding new settings) with local modifications (customizations of existing settings). The file would contain fragments from both versions, separated by clearly visible demarcation lines:

  <<<<<<< Updated upstream
  #ADMINTOOL_DEBUG=True
  ADMINTOOL_LOGIN_METHODS=google-oauth2 launchpad yahoo
  # choose from:
  #    shibboleth, locallogin, google-oauth2, google-plus, yahoo, amazon,
  #    docker, dropbox-oauth2, facebook, launchpad, linkedin-oauth2,
  #    meetup, twitter

  # Add any additional settings by prefixing them with ADMINTOOL_EXTRA_SETTINGS_
  # Example:
  #ADMINTOOL_EXTRA_SETTINGS_SOCIAL_AUTH_TWITTER_KEY=93randomClientId
  #ADMINTOOL_EXTRA_SETTINGS_SOCIAL_AUTH_TWITTER_SECRET=ev3nM0r3S3cr3tK3y
  =======
  ADMINTOOL_DEBUG=True
  >>>>>>> Stashed changes

As part of resolving the conflict, remove also the demarcation lines. In this example, the ADMINTOOL_DEBUG setting was modified (uncommented) in the local file, while right on the following line, new settings were added in the upstream. The correct resolution of thi confict is:

  ADMINTOOL_DEBUG=True
  ADMINTOOL_LOGIN_METHODS=google-oauth2 launchpad yahoo
  # choose from:
  #    shibboleth, locallogin, google-oauth2, google-plus, yahoo, amazon,
  #    docker, dropbox-oauth2, facebook, launchpad, linkedin-oauth2,
  #    meetup, twitter

  # Add any additional settings by prefixing them with ADMINTOOL_EXTRA_SETTINGS_
  # Example:
  #ADMINTOOL_EXTRA_SETTINGS_SOCIAL_AUTH_TWITTER_KEY=93randomClientId
  #ADMINTOOL_EXTRA_SETTINGS_SOCIAL_AUTH_TWITTER_SECRET=ev3nM0r3S3cr3tK3y

After editing the file, indicate to Git the conflict has been resolved: git reset
And dropped the stashed copy of the local modifications (git kept it because it ran into the conflict): git stash drop

Updating docker containers

After updating the files driving the tools:

Go into the respective directory (repeat for all of the three tools the updates apply to): cd admintool
Pull the updated container images: docker-compose pull
Restart the containers from updated images: docker-compose up -d
Optionally, watch the logs (leave with Ctrl-C): docker-compose logs -f

Troubleshooting

Especially when experimenting with the setup, it may happen that the tools get into an undefined and inoperative state.

For each of the tools, the following command would force-fully re-create the containers:

docker-compose stop
docker-compose rm --force
docker-compose up -d
docker-compose logs -f

The tools keep their external state in Docker volumes - with the bulk of the state being in the postgres database (for Admintool and Monitoring) and ElasticSearch (Metrics).

For the Admintool, the database volume can be deleted while the containers are stopped and removed so:

docker-compose stop
docker-compose rm --force
docker volume rm admintool_postgres-data
docker-compose up -d
docker-compose logs -f

Note that after this step, it would be necessary to run the admintool-setup.sh script again to initialize the new blank database.

Similarly, for the Monitoring tool, the steps to delete the database volume (while the containers are stopped and removed) would be:

docker-compose stop
docker-compose rm --force
docker volume rm icinga_postgres-data
docker-compose up -d
docker-compose logs -f

And again, it would be necessary to run the icinga-setup.sh script again to initialize the new blank database.

And likewise, for the Metrics tool, the steps to delete the elasticsearch data volume (while the containers are stopped and removed) would be:

docker-compose stop
docker-compose rm --force
docker volume rm elk_elasticsearch-data-6
docker-compose up -d
docker-compose logs -f

And in this case, it would be necessary to create the index from Kibana again on first access.

Note that each of the tools has multiple Docker storage volumes. The above commands include removing just the one which is the main datastore - so which when deleted likely removes the source of troubles.

To see the full list of volumes, run:

docker volume ls

If the above steps did not restore the tools to a clean state, you may also remove the additionla volumes (e.g., with logs,...) - please note the volumes can only be removed after stopping and removing the containers, as used in the above sequences.

Appendix: Google Login

To get the Google credential (key+secret) to use in the admintool, do the following in the Google Developer Console:

Start at http://console.developers.google.com/
Create a new project
From the main menu (top-left corner), select again APIs & Services and then Credentials
Configure the OAuth consent screen with how the application should be described to the user (at least, set Product name)
On the Credentials tab, create a new Credential as an OAuth Client ID for a web application
Add the Authorized redirect URI for your application - the URL should have the following form is (substitute your real hostname here):
```
  https://admin.example.org/accounts/complete/google-oauth2/
```
After saving, this gives you the Client ID and secret (use these as the GOOGLE_KEY and GOOGLE_SECRET)

Appendix: Google Maps API keys

The Admintool uses the Google Maps Javascript API to render the service location points on maps provided by Google.

As of June 22, 2016, the Google Maps Javascript API now requires an API key. Any new services deployed after this date (as determined by the domain registration date) require an API key. On such services, without an API key, the map view displays an error and the Javascript console shows messages indicating an API key is missing. However, even if your service is working without an API key, we strongly recommend to configure one, as Google may in the future decide to make this mandatory for all services. You can find more details about the API change itself in the original Google Announcement.

Further, as of July 2018, Google requires Billing to be configured on an account to allow increasing the quota from 1 page load per day.

However, according to the above announcement and the Maps Platform Pricing Sheet:

Google will apply a credit of $200 each month.
Which is equivalent to 28,000 map loads per month.

If you are concerned about incurring costs on your account, you can set the quota to a conservative value (like 900 page loads per day), making sure there would not be any charge actually billed.

To create a Google Maps API key:

Start at the Google Developer Console: http://console.developers.google.com/
Open the project you created earlier for configuring Google Login (or create a new one if you have not configured Google Login yet).
Enable APIs for you project:
- From the main menu (top-left corner), select the APIs & Services and then Library
- Search for Google Maps JavaScript API and Enable this for your project.
- Also search for Geocoding API and Enable this for your project.
- And also search for Places API and Enable this for your project.
We also highly recommend to set a quota for the Maps API calls (see the paragraph about Billing above).
- From the main menu (top-left corner), select the APIs & Services and then Dashboard
- From the list of APIs, select Maps JavaScript API and then select the Quotas tab.
- Edit the Map loads per day quota and set it to a suitable value - 500 will guarantee no charges under the current pricing regime (leaving also enough for the Geocoding API)
  - By default, the quota is unlimited - so uncheck the Unlimited checkbox first.
  - You may encounter a UI glitch where the text typed into the quota does not show up. It is, neverless recorded, so after typing in the desired value, Save - and check the result.
- And also set a quota for Geocoding API - again, navigate to the Dashboard and from there, pick the Geocoding API.
  - On the Quotas tab, scroll down to the Requests section (past Premium sections) and change Requests per day from Unlimited to about 500 - and Save.
- And also set a quota for Places API - again, navigate to the Dashboard and from there, pick the Places API.
  - On the Quotas tab, scroll down to the Requests section (past Premium sections) and change Requests per day from Unlimited to about 500 - and Save.
Now create credentials (the API key):
- From the main menu (again, top-left corner), select again APIs & Services and then Credentials
- From Create credentials, select API key
- This display the key you would be using - store the value.
- Contiue via the Restrict Key button to set restrictions on how your key can be used:
  - Pick a name for your credential - e.g., Google Maps API key
  - Under Application Restrictions, select HTTP referrers (websites) and enter the name of your website as the accepted referrer. This would be the hostname you entered as SITE_PUBLIC_HOSTNAME in your admintool.env prefixed with https:// so e.g.:
```
  https://admin.example.org
```
Under API Restrictions, choose which APIs this key can be used for. From the list of APIs enabled for your project, select Maps Javascript API and Geocoding API.
Save. And you can now configure this API key in the GOOGLE_API_KEY setting in admintool.env - and remember to update the Admintool containers with:
```
  cd etcbc-public/admintool
  docker-compose up -d
```

Appendix: Generating radsecproxy.conf from Admintool

The Admintool can generate radsecproxy.conf to be used by the NRO radsecproxy.

This would be done based on Institutional radius server information recorded in the application itself and based on TLR server configuration provided externally (via admintool.env)

The generated radsecproxy.conf configuration will:

Use a template preamble with generic settings
Define an client section for each Institutional radius server associated with at least one institution (and having the SP role).
Define a server section for each Institutional radius server associated with at least one institution (and having the IdP role).
Define a realm section for each institution's realm and map it to the institution's radius servers (having the IdP role).
Include a template blacklist.
Finish the blacklist with a "do not forward this countrie's requests upstream" rule, depending on the top-level-domain code for the deployment country.
And finally define the client, server and realm section for TLR servers, forwarding ALL requests not caught by the previous rules to the TLR servers.

Preparing radsecproxy.conf configuration in Admintool

There is a number of configuration variables in admintool.env that control how the configuration will be generated:

TLR_SERVERS specifies the list of TLR servers (short identifiers) - example: TLR_SERVERS=server1 server2
- This configuration closely follows the NRO_SERVERS setting used for Icinga configuration.
For each server, there should be a set of variables using the short server identifier as part of the name:
- TLR_SERVER_HOSTNAME_serverid: the hostname (or IP address) to use for server checks
- TLR_SERVER_SECRET_servererid: the Radius secret to use with the server
- TLR_SERVER_PORT_serverid: the Radius port number to use with this server (Optional, defaults to 1812)
- TLR_SERVER_STATUS_serverid: should be set to True if the server supports Radius Status checks (Optional, defaults to False)
RADSECPROXY_CONF_TLD: top-level domain of the deployment country (used to match all local realms)
Update Admintool to pick up the changed configuration:
```
  docker-compose up -d
```

As a final step, prepare an account that would be used to retrieve the configuration. The account created for fetching Icinga configuration can be reused for this purpose - if this is not desired, create a separate account following the same instructions as in the Preparing Icinga configuration in Admintool section.

To allow the account to access the radsecproxy.conf configuration:

Log into the Admintool admin interface at https://admin.example.org/admin/ as the administrator (with the username and password created earlier).
Select Users from the list of tables to administer.
Locate the account already created for fetching Icinga configuration and select to edit the account.
In the list of Available user permissions, select all three edumanage | Institution Server permissions and add them to the Chosen user permissions. (The permission to access the Institution Servers is internally used to represent as permission to access the radsecproxy.conf configuration - as it is at the same level of privilege, exposing server radius secrets).
- Select Save to store the permissions.

Now you should be able to access the monitoring configuration at https://admin.example.org/radsecproxyconf with this account. (And it is also accessible under the administrator account).

Entering data for monitoring configuration into the Admintool.

The following information needs to be entered into the Admintool in order to generate meaningful radsecproxy.conf configuration. Due to dependencies between different data objects, we recommend entering the data in the order given here.

This information is already included in the information entered for the monitoring configuration (assuming institutions' servers were included as well).

All of the data can be entered at https://admin.example.org/admin/

Add an Institution - select the NRO Realm, entity type and enter the English name of the Institution.
Add at least one Contact for the institution (not that Contacts can be reused across institutions).
Add an Institution Details object: select the Institution it belongs to, enter Address details and select at least one Contact.
Add an Institution Server: select the institution (or possibly more) the server belongs to, server type, a descriptive name and the host name, the Radius secret, and select whether the server responds to Status messages.
Add the Institution's Realms: enter the Realm name, select the Institution and select the institution's radius server that the NRO radius server should be proxying to.

Entering all of the above should be sufficient for the Admintool to generate monitoring checks for Icinga - both via the NRO servers and directly through the Institution's servers.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

Eduroam tools container-based deployment: Overall considerations

ChangeLog

Preliminaries - System considerations

Preliminaries - Docker

Preliminaries - Mail server

Eduroam ancillary tools: Basic setup

Deploying admintool

Installing proper SSL certificates for Admintool

Preparing Icinga configuration in Admintool

Entering data for monitoring configuration into the Admintool.

Deploying monitoring tools

Installing proper SSL certificates for Monitoring tools

Deploying metrics tools

Installing proper SSL certificates for Metrics tools

Feeding Radius logs from NRS servers (radsecproxy) into metrics tools (ELK)

Feeding Radius logs from NRS servers (freeradius) into metrics tools (ELK)

Accessing visualizations in metrics tools

Troubleshooting visualizations

Accessing the services

Configuring port numbers

Network access requirements

Updating deployed tools

Updating Docker files

Updating docker containers

Troubleshooting

Appendix: Google Login

Appendix: Google Maps API keys

Appendix: Generating radsecproxy.conf from Admintool

Preparing radsecproxy.conf configuration in Admintool

Entering data for monitoring configuration into the Admintool.

About

Releases

Packages

Languages

Name		Name	Last commit message	Last commit date
Latest commit History 176 Commits
admintool		admintool
elk		elk
filebeat-radius		filebeat-radius
filebeat-radsecproxy		filebeat-radsecproxy
icinga		icinga
Docker-compose-intro.md		Docker-compose-intro.md
Docker-intro.md		Docker-intro.md
Docker-setup.md		Docker-setup.md
Lab.md		Lab.md
README.md		README.md

REANNZ/etcbd-public

Folders and files

Latest commit

History

Repository files navigation

Eduroam tools container-based deployment: Overall considerations

ChangeLog

Preliminaries - System considerations

Preliminaries - Docker

Preliminaries - Mail server

Eduroam ancillary tools: Basic setup

Deploying admintool

Installing proper SSL certificates for Admintool

Preparing Icinga configuration in Admintool

Entering data for monitoring configuration into the Admintool.

Deploying monitoring tools

Installing proper SSL certificates for Monitoring tools

Deploying metrics tools

Installing proper SSL certificates for Metrics tools

Feeding Radius logs from NRS servers (radsecproxy) into metrics tools (ELK)

Feeding Radius logs from NRS servers (freeradius) into metrics tools (ELK)

Accessing visualizations in metrics tools

Troubleshooting visualizations

Accessing the services

Configuring port numbers

Network access requirements

Updating deployed tools

Updating Docker files

Updating docker containers

Troubleshooting

Appendix: Google Login

Appendix: Google Maps API keys

Appendix: Generating radsecproxy.conf from Admintool

Preparing radsecproxy.conf configuration in Admintool

Entering data for monitoring configuration into the Admintool.

About

Topics

Resources

Stars

Watchers

Forks

Releases

Packages 0

Languages

Packages