Open Stack

Hello everyone,

We're working on automatic evacuation in this approach:
1. We want to add extra information to nova servicegroup api regarding host health - https://review.openstack.org/#/c/137768/ 
2. We want to add new plugin to servicegroup mechanism which will use Pacemaker and provide best information about host health
3. Provide pacemaker configuration scripts for enable pacemaker to perform actual evacuation

This is basically what Russell spoke about, and while I agree that technically it's not required that nova would know about host health (as long as pacemaker will perform fencing on host down), I still think that nova should have this information for consistency of existing logic, for example:  https://bugs.launchpad.net/nova/+bug/1379292 . Nova service group is very misunderstood and often used as monitoring tool, even in nova (evacuation) or horizon (horizon presents message about host health based on servicegroup).

I think it would be good idea to actually provide quality information for such mechanisms to use.

Regards,
Michał Jastrzębski

> -----Original Message-----
> From: Russell Bryant [mailto:rbryant at redhat.com]
> Sent: Wednesday, December 3, 2014 7:52 PM
> To: Britten, Tyler; Jason Venner; Daniel P. Berrange
> Cc: Enterprise-wg at lists.openstack.org; Stefano Maffulli
> Subject: Re: [Win The Enterprise-wg] libvirtWatchdog status
> 
> On 12/03/2014 01:37 PM, Britten, Tyler wrote:
> > It seems like the main ask from the 'pets' side of the enterprise is
> > not instance monitoring/recovery, but hypervisor monitoring for
> > instance
> > recovery- KVM host fails, something is checking for a heartbeat, and
> > once that host is marked as offline, it would check the db for the
> > instances running on that host and schedule them to start on other
> > remaining hosts. Ovbiously this would require shared ephemeral storage
> > (NFS) or limit recovery to boot from volume instances.
> >
> > Am I offbase?
> 
> Nope, and there have been lots of discussions about how to best accomplish
> that.  Some think it should be built into Nova.  Others suggest other parts of
> OpenStack (existing or new components) it should be handled by.  Others
> (including me) suggest this should be (and can already be) handled by tools
> outside of OpenStack software.
> 
> Some references:
> 
> Latest thread from openstack-dev about this:
> 
> 
> http://lists.openstack.org/pipermail/openstack-dev/2014-
> October/048338.html
> 
> A blog post of mine about how I think it could be done from outside of
> OpenStack today:
> 
> 
> http://blog.russellbryant.net/2014/10/15/openstack-instance-ha-proposal/
> 
> Unfortunately I haven't been able to follow that up with next steps of
> prototyping and testing to prove it out in more detail.
> 
> --
> Russell Bryant
> 
> _______________________________________________
> Enterprise-wg mailing list
> Enterprise-wg at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/enterprise-wg