← Back to team overview

nova team mailing list archive

Re: Got nova running in a single node as per the docs


Hey Manuel,

I sent this message last night which details the flags you need to specify on the other machines.  It looks like it didn't make it through, so here it is again:

Sorry Manuel,  those docs are a bit dated.  NGINX isn't required anymore with the new objectstore. I use the following script to install and run the current version of nova:


It runs nova-volume in fake mode, but it shouldn't be too hard to hack/edit to make volumes work.  Multiple nodes  is pretty straightforward if your router supports host-specified vlans.  You will have to set up your router to allow one or more host managed vlans.  The easiest set up is to run everything but compute and volume on one host, then run either dedicated compute and volume hosts, or hosts that run both compute and volume.  You will need to specify the following flags for nova-compute and nova-volume
--vlan_start=<vlans start number>
--redis_host=<main host ip> (or --ldap_url if you are using real ldap)
--rabbit_host=<main host ip>
--s3_host=<main host ip>
--ec2_url=http://<main host ip>:8773/services/Cloud
--ldap_url=ldap://<main host ip>

You'll also need to install a real sql database as sqlite only works on one box.  An easy setup is to install mysql, edit my.conf to listen on the public ip.  Create a database, user and password and pass the following flag into all
of the workers:

remember, you can stick your flags in a flagfile and pass --flagfile=/path/to/flagfile on the command line.

Good luck,

On Sep 16, 2010, at 1:06 PM, Manuel Amador (Rudd-O) wrote:

> OK.  So right now I have a one-machine cloud.
> Now let's assume that I want another machine to join the cloud.
> What steps should I run there?
> El Thursday, September 16, 2010, Joshua McKenty escribió:
>> Rudd-O,
>> An awesome report - looks like there are four or five critical updates to
>> the instructions, two minor bugs and a big gap in the user docs (e.g., what
>> next?)
>> Many folks are still using euca-tools or some ec2 tools to send API
>> requests to a running nova cloud - you'll need to use the nova-admin
>> binary to set up user accounts and download credentials to use with those
>> tools. (I was pretty sure there was a doc on that somewhere...)
>> Maybe Devin or a racker can pipe up about the status of the new API
>> endpoint - is it ready enough to use? (If so there, are a bunch of clients
>> already built for that API, including android and iphone apps, etc.)
>> If the objectstore is natively running on :3333, then perhaps we don't need
>> the nginx hack anymore? (Sorry, haven't looked at that code in a while.)
>> Otherwise, you'll want to change your objectstore flagfile to run on port
>> 3334, and let nginx proxy to it when appropriate.
>> You should be able to get play-by-play help in the IRC channel, once folks
>> are awake.
>> Joshua
>> On Thu, Sep 16, 2010 at 7:24 AM, Manuel Amador (Rudd-O)
>> <rudd-o@xxxxxxxxxx>wrote:
>>> in here: http://nova.openstack.org/getting.started.html
>>> Now what?  How do I tell it to create VMs?  Start VMs?  Stop VMs?  What
>>> do I
>>> install in other nodes and How do I make those other nodes join that
>>> cloud?
>>> These are my notes --- corrections on the install docs:
>>> ----------------------------------------------
>>> http://nova.openstack.org/getting.started.html
>>> KVM:Maverick-x86_64
>>> address
>>> user name rudd-o
>>> password (short one)
>>> I built redis from 2.0.1 stable source downloaded from
>>> http://code.google.com/p/redis/downloads/detail?name=redis-2.0.1.tar.gz&c
>>> an=2&q=
>>> I then ran openstack/nova/nova/auth/slap.sh to set up slapd
>>> I had to install python-tornado and python-gflags too
>>> The instructions were wrong, there is no KVM (hence no libvirtd group) in
>>> the
>>> cloud controller
>>> I created the .pth file too
>>> To run nova-api, I had to install python-daemon too
>>> To run nova-api, I had to install python-carrot too
>>> To run nova-api, I had to install python-ipy -- without it, import of
>>> class nova.network.manager.VlanManager fails -- the utils.import_object
>>> function makes the mistake of hiding the true source of the import error
>>> by overriding
>>> importerror with nova.exceptions.NotFound, which misleads the user to
>>> think that VlanManager does not exist, when in fact it is a dependency
>>> problem that
>>> causes it to "not exist".
>>> I had to install python-sqlalchemy too, to run nova-compute
>>> All the daemons i started with --nodaemon
>>> NGINX as configured as per the instructions just cnanot be run because it
>>> causes nova-objectstore to simply fail to start up.
>>> 2010-09-15 19:48:05-0700 [-] twisted.internet.error.CannotListenError:
>>> Couldn't listen on any:3333: [Errno 98] Address already in use.
>>> Had to stop nginx because of that.
>>> _______________________________________________
>>> Mailing list: https://launchpad.net/~nova
>>> Post to     : nova@xxxxxxxxxxxxxxxxxxx
>>> Unsubscribe : https://launchpad.net/~nova
>>> More help   : https://help.launchpad.net/ListHelp
> _______________________________________________
> Mailing list: https://launchpad.net/~nova
> Post to     : nova@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~nova
> More help   : https://help.launchpad.net/ListHelp

Follow ups