vCPE use case testing

Owned by Shankaranarayanan Puzhavakath Narayanan
Last updated: Aug 15, 2018 by Matthew DavisVersion comment
10 min read

Here's the list of requirements that need to be met at each of the components for the homing call from SO to succeed for the vCPE use case. 

Requirements from the downstream dependencies

AAI

Data needed

  • vGMuxInfra (run the vcpe flows that create the vGMuxInfra)

  • list of Cloud regions, one of which should have the vGMuxInfra

MultiCloud

Data needed

Policy

Policy models to be created:

Policies to be uploaded: 

  • vG demand

vGMuxInfra demand

  • objective function (distance) 

  • colocation constraint 

HAS



OSDF 

Requirements for Functional Features



HPA

  • upload HPA policies into policy framework



Capacity checks

  • upload Capacity check policy into policy framework



Sequence of actions (how to..)

  • Deploy OOF

    • HEAT: 

      1. Go to the OOF VM

      2. Run /opt/oof_vm_init.sh - this will pull the docker images and install OSDF + HAS, and perform the required configurations to pass the healthcheck. 

    • OOM:



  • populate the cloud region data into AAI

  • upload the policies into policy framework 

Run the vcpe flows that create the vGMuxInfra

vCPE use case standalone testing on a single machine

Goal: Running the vCPE use case of the Beijing release of ONAP in a local setup, with minimal hardware and software requirements.
The work should be done in the ONAP MUSIC and ONAP HAS repositories with tag 2.0.0-ONAP.

Prerequisites



  1. Install Docker CE, an example tutorial can be found here.

  2. Make local copies of the HAS and MUSIC repos

    HAS and MUSIC install

    $ cd "your workspace" $ git clone https://gerrit.onap.org/r/music $ cd music $ git checkout 2.0.0-ONAP $ cd ../ $ git clone https://gerrit.onap.org/r/optf/has $ cd has $ git checkout 2.0.0-ONAP

  3. (Optional) Install Insomnia REST API client for handling REST requests. Some prepared requests and variables are available can be imported to Insomnia. Importable to Insomnia: 

  4. (Optional) For advanced debugging of the inter-component communication use Wireshark

Installation and Configuration

  1. Build and start A&AI simulator

    $ cd ~/has/conductor/conductor/tests/functional/simulators $ ./build_aaisim.sh $ ./run_aaisim.sh

    Note: DNS name resolution issue may arise during building the simulator, because it needs the **web** Python package. TBD: Provide more detailed instructions. How does one there is an issues? what can one do about it?

  2. Verify A&AI Simulator (Hint: in examples, 0.0.0.0 is IPAddress of A&AI container as per $ docker container ls)

    1. $ docker container ls

    2. expected result: image names should include: aaisim

    3. $ curl http://0.0.0.0:8081/healthcheck

    4. expected result: {"status": "success"}

    5. $ curl http://0.0.0.0:8081/aai/v13/cloud-infrastructure/cloud-regions/

    6. expected result: {"cloud-region": [{"group-assignments": ...much more...

  3. Build and start MUSIC

    1. Configure access to MUSIC:

    2. Edit music/distribution/dockermusic/music.sh to change user/password:

      ```text -CASS_USERNAME=cassandra1 -CASS_PASSWORD=cassandra1 +CASS_USERNAME=conductor +CASS_PASSWORD=c0nduct0r ```

    3. Edit music/distribution/dockermusic/properties/music.properties to change user/password:

    4. Build and start MUSIC containers

    5. Verify MUSIC components

      1. $ docker container ls

      2. expected result: image names should include: music-tomcat, music-zk, music-war, music-db

  4. Build and Start HAS

    1. Configure HAS

      1. Edit has/conductor.conf

      2. search for [music_api]



      3. Set the replication_factor to 1. (In the local installation MUSIC runs in a single instance, if this value is higher MUSIC requests with atomic consistency will fail with error message "Exception thrown while doing the critical put, check sanctity of the row/conditions:\nnull")

      4. Search for [aai]



      5. Hint: find IPAddress of containers via:

        1. $ docker container ls

        2. look at first term in PORTS column

  5. Onboard the HAS "conductor" application into MUSIC, 

    1. E.g.:

    2. expected result: {"Generated AID":"2895db7e-2746-426a-822f-80c367161c65","Success":"Your application conductor has been onboarded with MUSIC."}

  6. Start the HAS containers

    1. $ cd ~/has

    2. $ ./run-dockers.sh

  7. Verify HAS components

    1. $ docker container ls

    2. expected result: image names should include: data, reservation, solver, api, controller

  8. At this moment a $ docker ps or $ docker container ls commands should have an output similar to the following

    1. Example output

Running

  1. Send vCPE homing template to HAS, E.g.

    1. $ curl -X POST --user admin1:plan.15 -H 'Content-Type: application/json' \
      -d @homing.json localhost:8091/v1/plans

    2. Hint: homing.json may be created by conjoining 

      where HAS Homing Specification (R') = https://lf-onap.atlassian.net/wiki/display/DW/vCPE+Homing+Use+Case

    3. Hint: for username:password (--user above) see ~/conductor.conf, conductor_api section

  2. RESOLVED: ISSUE_1: At this point a request fails with an error and the following message is returned for the homing request:

    1. After a 5 mins long timeout, the HAS api component replies back with error:

    3. The underlying communication can be observed by capturing network traffic on the docker bridges by Wireshark. The communication is stuck (which causes the timeout at the api component) after a faulty MUSIC reply:

      1. Request to MUSIC sent by the controller: PUT /MUSIC/rest/v2/keyspaces/conductor_rpc/tables/controller/rows?id=<<Some UUID>>

        1. This request is a database primitive of the MUSIC API which modifies a row identified by <<Some UUID>> in table *controller*  of keyspace *conductor_rpc*. The data of the new database entry is delivered in the HTTP request body (not included here), which can be captured by Wireshark observing the communication between the controller HAS component and MUSIC.

      2. MUSIC's reply: "Exception thrown while doing the critical put, check sanctity of the row/conditions:\nnull"

      3. The produced HTTP stream between `controller` and MUSIC captured with Wireshark:

  3. ISSUE_2: Controller - MUSIC interaction fails with "Customer ID not specified for demand vGMuxInfra". This error message can be accessed by capturing HTTP traffic on the docker bridge created for the HAS components. Even though a HTTP 200 OK message is sent back to the homing template vCPE request initiator's side (either the Insomnia REST client or the curl command), the placement does not finish successfully.

Contact

Please if you have anything to add to or ask about this tutorial, contact me: balazs.nemeth__AT__tmit.bme.hu