Open Stack

Chris, You are definitely on to something here.  When I create the instances individually this condition does NOT occur.  I confirmed this by creating 20 instances with openstack server create all on the same host. I then set the host to disabled in horizon and used the migrate host button.  This time the scheduler worked as expected, migrated one at a time as specified in the controller max_concurrent_live_migrations=1 and queued the rest until they all completed and the host was empty. 

Steven Searles



-----Original Message-----
From: Steven D. Searles [mailto:SSearles at zimcom.net] 
Sent: Wednesday, September 20, 2017 12:16 AM
To: Chris Friesen <chris.friesen at windriver.com>; openstack at lists.openstack.org
Subject: Re: [Openstack] Pike NOVA Disable and Live Migrate all instances.

I did.  I will spawn a few singles and see if it does the same thing. 


Steven Searles




-----Original Message-----
From: Chris Friesen [mailto:chris.friesen at windriver.com]
Sent: Tuesday, September 19, 2017 11:17 PM
To: openstack at lists.openstack.org
Subject: Re: [Openstack] Pike NOVA Disable and Live Migrate all instances.

On 09/19/2017 05:21 PM, Steven D. Searles wrote:
> Hello everyone and thanks in advance.  I have Openstack Pike
> (KVM,FC-SAN/Cinder) installed in our lab for testing before upgrade 
> and am seeing a possible issue with disabling a host and live 
> migrating the instances off via the horizon interface. I can migrate 
> the instances individually via the Openstack client without issue. It 
> looks like I might be missing something relating to concurrent jobs in 
> my nova config? Interestingly enough when a migrate host is attempted via horizon they all fail.  Migrating a single instance through the horizon
> interface does function.   Below is what I am seeing in my scheduler log on the
> controller when trying to live migrate all instances from a disabled 
> host. I believe the last line to be the obvious issue but I can not 
> find a nova variable that seems to relate to this.  Can anyone point me in the right direction?

There's a default limit of 1 outgoing live migration per compute node.  I don't think that's the whole issue here though.

> *2017-09-19 19:02:30.588 19741 DEBUG nova.scheduler.filter_scheduler 
> [req-4268ea83-0657-40cc-961b-f0ae9fb3019e
> 385c60230b3f49da930dda4d089eda6b
> 723aa12337a44f818b6d1e1a59f16e49 - default default] There are 1 hosts 
> available but 10 instances requested to build. select_destinations
> /usr/lib/python2.7/dist-packages/nova/scheduler/filter_scheduler.py:10
> 1*

It's unclear to me why it's trying to schedule 10 instances all at once.  Did you originally create all the instances as part of a single boot request?

Chris

_______________________________________________
Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to     : openstack at lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
_______________________________________________
Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to     : openstack at lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack