Need to increase size of RAM for virtual machines to deploy_bonding_neutron_vlan

[Alexandr Kostrikov]

During swarm test run I found that job has been failed.

After revert I found that virtual machine with controller is under high load (even show as offline).

Before ssh to node:
[root@nailgun ~]# fuel nodes
id | status | name | cluster | ip | mac | roles | pending_roles | online | group_id
---|--------|---------------------|---------|-------------|-------------------|------------|---------------|--------|---------
5 | ready | slave-05_cinder | 1 | 10.109.12.7 | 64:16:93:f3:ce:8c | cinder | | True | 1
2 | ready | slave-01_controller | 1 | 10.109.12.3 | 64:32:ce:3a:b7:69 | controller | | False | 1
3 | ready | slave-03_controller | 1 | 10.109.12.5 | 64:dd:45:92:d6:3f | controller | | True | 1
1 | ready | slave-02_controller | 1 | 10.109.12.4 | 64:00:5b:2f:23:3c | controller | | True | 1
4 | ready | slave-04_compute | 1 | 10.109.12.6 | 64:9a:0b:f6:27:26 | compute | | True | 1

When I did ssh to node - it took a long time and htop shown high load.
After some time it has been shown as online.
[root@nailgun ~]# fuel nodes
id | status | name | cluster | ip | mac | roles | pending_roles | online | group_id
---|--------|---------------------|---------|-------------|-------------------|------------|---------------|--------|---------
5 | ready | slave-05_cinder | 1 | 10.109.12.7 | 64:16:93:f3:ce:8c | cinder | | True | 1
2 | ready | slave-01_controller | 1 | 10.109.12.3 | 64:32:ce:3a:b7:69 | controller | | True | 1
3 | ready | slave-03_controller | 1 | 10.109.12.5 | 64:dd:45:92:d6:3f | controller | | True | 1
1 | ready | slave-02_controller | 1 | 10.109.12.4 | 64:00:5b:2f:23:3c | controller | | True | 1
4 | ready | slave-04_compute | 1 | 10.109.12.6 | 64:9a:0b:f6:27:26 | compute | | True | 1

[Alexandr Kostrikov (akostrikov-mirantis)]

That affected only one test, removing swarm-blocker.

[Alexandr Kostrikov (akostrikov-mirantis)]

I have run following command: `ps -eLo pid,ppid,rss,cmd | sort -rnk3`
Output is at http://paste.openstack.org/show/508562/

Please, nova team, prove that such resource consumption by nova-api is acceptable and we should increase memory on that job. Or it is abnormal and nova-api should be tweaked.

[Alexandr Kostrikov (akostrikov-mirantis)]

The failure is reproducible, feel free to request for env at https://product-ci.infra.mirantis.net/job/9.0.system_test.ubuntu.bonding_ha/

[Roman Podoliaka (rpodolyaka)]

Alenxandr, I'd say anything below ~200 Mb is perfectly normal - this is what you see after nova-api imports all the modules used. Short spikes of memory usage above 200 Mb is also ok, as long as it goes back to the previous level after processing of incoming requests.

[Andrey Sledzinskiy (asledzinskiy)]

Set to won't fix for 9.0, we won't be able to make a decision about updating ram before HCF

[Fuel Devops McRobotson (fuel-devops-robot)]

Fix proposed to branch: master
Change author: Alexandr Kostrikov <email address hidden>
Review: https://review.fuel-infra.org/21936

[Fuel Devops McRobotson (fuel-devops-robot)]

Reviewed: https://review.fuel-infra.org/21936
Submitter: Lesya Novaselskaya <email address hidden>
Branch: master

Commit: 842b387bd5c571571822f2d9890cf6934ba00e02
Author: Alexandr Kostrikov <email address hidden>
Date: Fri Jun 10 11:22:35 2016

Increase memory for bonding tests

There is lack of resources on controllers
to check ha with bonding.

Change-Id: Icd0eefce98967ce09bc4d5cba501986659bb1ede
Closes-bug: #1587466