Open Stack

Hi all,

This is not just a documentation issue, if it was I would definitely pay for tech writers to resolve - and will do that once we get the issues fixed there are a number of them which we have documented here:

https://etherpad.openstack.org/p/vC6iMwOWft - one of our Intel engineers is working this list,it includes issues related to HA developers and operators are aware of.

Here is our current status:
Right now we're working to fix Nova and Cinder issues. In Nova we want to make nova-scheduler more reliable providing retries if a service dies during request processing. In Cinder we're fixing similar problems in a more flexible way - implementing persistence of workflows to be able to resume them after service crash or shutdown. This should help on the problem operators are complaining all the time - Cinder resources left in unresolved states.

We should be clear on HA meaning that the services are online (typically expect the services to respond with 99.99% availability- no more than 52mins of downtime a year).  A service not responding is considered down, and a service that loses transactions is down.

For instance Heat services today are not built in a resilient fashion, and therefore if a single node with the Heat service running dies it does not send the work to another node for that stack....  This is not HA.

We are committed to engineer in the community solutions to resolve these, and we will definitely help with docs too.

-Das


***enohp opyt eht morf tnes***

On Jan 13, 2015, at 9:57 PM, Stefano Maffulli <stefano at openstack.org<mailto:stefano at openstack.org>> wrote:

On Thu, 2015-01-08 at 20:04 +0000, Barrett, Carol L wrote:
WTE Team – Enterprises have HA requirements and several of the WTE
teams are working on different angles of this.

Are any of you involved in with the HA Guide update underway? If not,
should we be?

Thanks for bringing this up Carol. I have no doubt that people in this
group should be following the HA Guide closely because most of the HA
capabilities that enterprises need are already available in OpenStack,
today, only they're documented at sub-optimal levels and not emphasized
in communication.

To your question I'd add:

- Who is willing to put resources to read the guide, test its validity
and provide feedback to the writers?

- Who has tech writers to contribute to the writing effort?

I believe that being generically 'involved' is not enough anymore and
there needs to be a clear 'commitment'. Matt and Sriram have only
started and I'm sure they'd appreciate more people.

Cheers,
stef


_______________________________________________
Enterprise-wg mailing list
Enterprise-wg at lists.openstack.org<mailto:Enterprise-wg at lists.openstack.org>
http://lists.openstack.org/cgi-bin/mailman/listinfo/enterprise-wg