Ask Your Question
0

Problem with launch big image instance (1 GB+) on Ocata

asked 2017-04-30 11:58:15 -0500

vl_dorfan gravatar image

Ocata release Two strong nodes setup controller and compute. Most services running on controller. There no problem to launch cirros or any small flavor like fedora( ~250 MB) The compute(M4 IBM server) have 128 GB memory and 1.4 TB disk available for instances, CPU `24 physical cores(2 sockets). The limit defined as 100 vCPU... All neutron agents up(ml2 , linux-bridge etc) and running same for nova services , the controller show hypervisor up and running compute node appear properly, keystone ok , glance service also.- everything run according to manual(RDO) without any warnings in log files. The only one problem not enough connections in Rabbitmq fixed , increased from 214 to 4096. Deploying any big -more then 1 GB image VM failed with "No valid host was found. There are not enough hosts available. Code 500" and lot of python exceptions... Seems like same unknown limitation s in nova.conf. Any suggestion more than welcomed.

edit retag flag offensive close merge delete

Comments

What do you mean by a 1GB image VM? Is the image larger than 1GB or the instance?

The nova scheduler log should tell you if the placement service was unable to find a fitting compute host or if the filters rejected all hosts, and which filter. You may have to switch on debug logging.

Bernd Bausch gravatar imageBernd Bausch ( 2017-05-01 10:08:49 -0500 )edit

1 answer

Sort by ยป oldest newest most voted
0

answered 2017-05-03 00:39:20 -0500

vl_dorfan gravatar image

Mentioned qcow file used to deploy instance on Ocata.File copied to the local server and then used the command to create the image like:"OpenStack image create --disk-format qcow2 --container-format bare --public --file /tmp/NX-14.7.10-03-disk1.qcow2 NX1114" I used my own flavor" "big" defined as 8*VCPU *GB RAM and 80 GB HD Firstly I trying to deploy my own qcow file ~6GB(come from KVM) and failed.Then I decide to deploy some standard image downloaded from public location Win2008 for example on same flavor and faile on "not enough hosts available" I set debug on True in my nova.conf but there no more additional info recevied...

edit flag offensive delete link more

Comments

Which hypervisor are you using?

eblock gravatar imageeblock ( 2017-05-03 04:17:39 -0500 )edit

KVM according to H/W features

vl_dorfan gravatar imagevl_dorfan ( 2017-05-03 05:01:14 -0500 )edit

What do you see in the scheduler log?

Bernd Bausch gravatar imageBernd Bausch ( 2017-05-03 05:08:58 -0500 )edit

2017-05-03 10:06:16.845 9489 INFO nova.scheduler.host_manager [req-8c919d95-de2b-44fb-b0e0-8244972b0742 - - - - -] Successfully synced instances from host 'compute1'. 2017-05-03 10:08:18.395 9489 INFO nova.scheduler.host_manager [req-c2c87c2d-8927-4ff0-8133-5e6fef25fbfc - - - - -] Successfully synce

vl_dorfan gravatar imagevl_dorfan ( 2017-05-03 05:46:56 -0500 )edit

The comment field only allows 255 characters. Try to find relevant log entries and post them in the question or paste.openstack.org.

Bernd Bausch gravatar imageBernd Bausch ( 2017-05-03 11:15:18 -0500 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Get to know Ask OpenStack

Resources for moderators

Question Tools

2 followers

Stats

Asked: 2017-04-30 11:58:15 -0500

Seen: 367 times

Last updated: May 03 '17