Questions & Answers

Ask a Question

Installing PNDA on a Machine with Devstack

0 votes
asked Apr 27 by kingpinH (320 points)
The guide is taking me all over the place wrt setting up pnda. I have a machine with devstack where I can create networks and spawn VMs. What should be the next steps to bring up pnda?

1 Answer

0 votes
answered Apr 27 by James Clarke (1,630 points)
PNDA has been successfully deployed using Devstack. You want to start here: https://github.com/pndaproject/pnda-guide/tree/master/provisioning, and work carefully through "Platform requirements", "Building PNDA", "Getting started with Heat", "Creating images for use with Heat templates" and "Using the PNDA Heat templates". Feel free to ask specific questions on any steps and we'll do our best to help out.
commented Apr 27 by yusong0926 (240 points)
edited Apr 27 by yusong0926
I thought it cannot be deployed to a single machine server.
commented Apr 28 by James Clarke (1,630 points)
Devstack is a distribution of Openstack, that will need to provide a minimum of the 4/5 VMs for a pico flavour of PNDA.
commented Apr 28 by yusong0926 (240 points)
so basically we are able to deploy it to a single node server with openstack on it which can create multiple vms for pnda nodes.
commented May 2 by kingpinH (320 points)
Quite honestly, the guide is written in a very circuitous way. There are too many forward references which makes it hard to go step by step. The formatting is unreadable at many places. While I would like to contribute to a better guide once I get my setup up, I am just hitting walls after walls trying to bring it up.

In Openstack requirements, suddenly a cluster is talked about. What cluster is this? Do I have to create a neutron network first? What is the prerequisite? I'll ask many different/specific questions as directed.
commented May 2 by trsmith2 (1,740 points)
If you could provide some specific references - i.e. to pages / line numbers - to the areas you're having problems with we'd be happy to make improvements.

Note that GitHub changed their support for Markdown after we released 3.4 so the tables in the platform requirements now appear badly formatted. You can refer to the develop branch where this is fixed.
commented May 2 by yusong0926 (240 points)
just want to be clear if pnda can be installed in a vm of a single node server.
commented May 2 by trsmith2 (1,740 points)
It does not matter how many server nodes are used to create OpenStack providing the minimum number of CPUs, memory and storage are made available.

The smallest profiles we have create 6 VMs on OpenStack.
commented May 2 by yusong0926 (240 points)
ok , got it. thanks! I agreed with kingpinH, the guide and repo need to be cleaned and improved. I guess everyone can provide help on the github.
just one place I found might be something missing yesterday when I was trying to deploy the example applications. In the example application repo, it seems lots of applications are missing upstart.conf and yarn-kill.py which seems are required when deploying the package. I just simply copy them into the tar.gz file to get it fixed.
commented May 2 by trsmith2 (1,740 points)
Here are the files you're looking for -

https://github.com/pndaproject/example-applications/blob/master/spark-streaming/app-package/src/main/resources/sparkStreaming/example/upstart.conf
https://github.com/pndaproject/example-applications/blob/master/kafka-spark-opentsdb/src/universal/sparkStreaming/sampleKSOapp/upstart.conf
https://github.com/pndaproject/example-applications/blob/master/h2o-launcher/app-package/src/main/resources/sparkStreaming/h2o/upstart.conf
...