Open Stack

Hi Armando, 

We are working on a custom implementation for nova-ha. It
is going to be a PHP daemon which decides whether a compute-node is down by
ping results of a combination of the node itself and the instances running
on it. 

The algorithm which decides the appriopriate actions is done, now
we are working on how to start an instance on another node. 

This (in my
view) can be done with the following steps; 

1. Start the instance in
virsh
 * Add the NWfilter by the XML file
 * Add the instance information
by the (qemu) XML file 
 * Start the instance
 2. Update IPtables
 * Add
the standard instance firewall rules
 * Add the extra firewall rules
provided by the instance's security group
 3. Add floating IP's
 * Add the
instance's floating IP's with 'ip addr add'
 * Add the IPtables rules to
forward the IP to the instance
 * Arping the floating IP's (?)
 4. Update
the nova database
 * Make sure the controller knows that the instance is
running on a different compute node by updating the instances table 

These
steps should be enough to bring an instance up on another node, or am I
missing something? 

The first step is actually quite easy if your running
shared storage, the only thing we need to make sure is that the libvirt xml
files are also on the shared storage space to make sure other nodes can add
the XML files. 

Step 2 and 3 is scriptable, allthough it would be nice if
there is any way to trigger the nova-api to rebuild the nova-networking for
an instance. I have tried rebooting an instance but that does not trigger
it. 
 Anyone has any idea if this is possible? 

Thanks! 

ps. Armando,
when we are done and successfully implemented our patch, I will leave a
note on the blueprint with our findings.  
---
Met vriendelijke
groet,
Gerrit Tamboer

On Mon, 30 Jan 2012 15:12:01 +0000, Armando
Migliaccio  wrote:   

As for issue 1, have you thought of implementing the
stop functionality as a combination of Snapshot+Terminate? 

As for issue
2, there is a blueprint [1] to deal with these sorts of scenarios.
Unfortunately that's not going to help you in the immediate term because
the effort hasn't started yet (for lack of time more than anything else).


Hope this help! 

Armando 

FROM:
openstack-operators-bounces at lists.openstack.org
[mailto:openstack-operators-bounces at lists.openstack.org] ON BEHALF OF
Gerrit Tamboer
SENT: 29 January 2012 09:38
TO:
openstack-operators at lists.openstack.org
SUBJECT: Re: [Openstack-operators]
A couple of questions regarding Nova   

Hi! 

Unfortunately, no reply
yet.
Does anyone have an idea how to solve these issues? 

Especially the
second question is a pain :) 

Regards, 

Gerrit 

On Fri, 27 Jan 2012
09:25:41 +0100, Gerrit Tamboer wrote:  

Hi! 

We are currently testing the
Openstack Nova product, but we have 2 problems/questions which we haven't
find a decent solution for yet. 

1. In The Nova-api v1.1 there is no
option to Stop and Start an instance, only Reboot and Terminate. I know
that these options are available in v2.0 which is in Beta right
now.
Because we would like to work with the stable version of the Nova-Api
we are not planning to move to v2.0 for now. 
We are wondering if anyone
has managed to find a solution for this issue? Is there a 'clean' way to
implement and use a Stop and Start function in the API? We can of course
make an exception by using custom shell scripts for the stop and start
function (using virsh destroy/start) but this is kinda nasty ;)..  

2. On
our Nova-Compute machines we have shared storage using an iSCSI target
which gives us the ability to use Live Migration. This is working
perfectly! The only problem we are facing at this moment is that we want to
be able to start an instance on a different Compute Node when the original
Compute Node goes down. Is there an easy way or (even better) an API call
which can be used for this? The Live Migration function does not work when
the original Nova Compute node is down. We are currently planning to
implement our own patch which detects a compute node going down and
automatically starts the instances on different compute nodes using various
algorithms. Only the last part (starting an instance on a different node)
is something we need to figure out, hopefully you guys can help! :)


Regards, 

Gerrit     

Links:
------
[1]
https://blueprints.launchpad.net/nova/+spec/guest-ha
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20120131/0ce0d617/attachment-0002.html>