| 2014-05-21 15:14:45 |
Timur Nurlygayanov |
description |
Environment:
MOS 4.1a, OpenStack Havanna, CentOS 6.4, Neutron with GRE, 1 controller node and 4 compute nodes.
Steps To Reproduce:
1. Create 200 VMs with Cirros Image.
2. Delete 100+ VMs with one request (in Horizon dashboard)
Observed Result:
Part of VMs will be successfully removed, but when we select the large group of servers, we can see the error message in Horizon and the traceback in nova.log and neutron.log
The root of this problem in the limit on requests/seconds for Neutron API and we can fix this problem - we should just increase the number of Neutron workers: need to set 'api_workers = 10' in /etc/neutron/neutron.conf.
Please, see for more detailed information:
https://bugs.launchpad.net/neutron/+bug/1321626
Note:
This issue is important for enterprise-deployment, where users can create more that 202 VMs simultaneously.
Of course, we can not predict the correct number of workers for any deployments, but we can set the basic configuration with 10 or 20 workers to cover many typical installations. Also, we can add a few words in the documentation about this problem. |
Looks like this issue is already fixed in Icehouse: https://review.openstack.org/#/c/58017/
Environment:
MOS 4.1a, OpenStack Havanna, CentOS 6.4, Neutron with GRE, 1 controller node and 4 compute nodes.
Steps To Reproduce:
1. Create 200 VMs with Cirros Image.
2. Delete 100+ VMs with one request (in Horizon dashboard)
Observed Result:
Part of VMs will be successfully removed, but when we select the large group of servers, we can see the error message in Horizon and the traceback in nova.log and neutron.log
The root of this problem in the limit on requests/seconds for Neutron API and we can fix this problem - we should just increase the number of Neutron workers: need to set 'api_workers = 10' in /etc/neutron/neutron.conf.
Please, see for more detailed information:
https://bugs.launchpad.net/neutron/+bug/1321626
Note:
This issue is important for enterprise-deployment, where users can create more that 202 VMs simultaneously.
Of course, we can not predict the correct number of workers for any deployments, but we can set the basic configuration with 10 or 20 workers to cover many typical installations. Also, we can add a few words in the documentation about this problem. |
|
