With config-download deployment, OVS-DPDK is enabled after NetworkDeployment

Bug #1807959 reported by Saravanan KR
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
tripleo
Fix Released
Medium
Saravanan KR

Bug Description

With config-download deployment, NetworkDeployment resource runs before the EnableDpdkDeployment, which is resulting in vswitchd error:
-----------------------------------
2018-12-09T16:08:26.993Z|00058|netdev|WARN|could not create netdev dpdk1 of unknown type dpdk
2018-12-09T16:08:26.993Z|00059|bridge|WARN|could not open network device dpdk1 (Address family not supported by protocol)
-----------------------------------

This happens because os-net-config runs before enabling DPDK in OvS. And later when DPDK is enabled, it continues to work event for the existing ports. But the error logs are misleading and it has changed the behavior from queens.

Ensure "EnableDpdkDeployment" runs before "NetworkDeployment" resource.

Revision history for this message
Saravanan KR (skramaja) wrote :

[root@undercloud overcloud]# cat group_vars/ComputeOvsDpdk
ComputeOvsDpdk_pre_deployments:
  - EnableDpdkDeployment
  - NetworkDeployment
  - ComputeOvsDpdkUpgradeInitDeployment
  - ComputeOvsDpdkDeployment
  - ComputeOvsDpdkHostsDeployment
  - ComputeOvsDpdkAllNodesDeployment
  - ComputeOvsDpdkAllNodesValidationDeployment
  - ComputeOvsDpdkArtifactsDeploy
  - ComputeOvsDpdkHostPrepDeployment

ComputeOvsDpdk_post_deployments: []
[root@undercloud overcloud]# cat group_vars/Controller
Controller_pre_deployments:
  - NetworkDeployment
  - ControllerUpgradeInitDeployment
  - ControllerDeployment
  - ControllerHostsDeployment
  - ControllerAllNodesDeployment
  - ControllerAllNodesValidationDeployment
  - ControllerArtifactsDeploy
  - ControllerHostPrepDeployment

Controller_post_deployments: []
---------------------------------------------------------

lookup of vars between multiple roles is not maintaining the order. In the above deployment, since NetworkDeployment is present on both roles, NetworkDeployment is deployed first on both the roles before the EnableDpdkDeployment, which is resulting in the order change.

Changed in tripleo:
assignee: nobody → Saravanan KR (skramaja)
importance: Undecided → Medium
status: New → Triaged
milestone: none → stein-2
Changed in tripleo:
milestone: stein-2 → stein-3
Revision history for this message
Saravanan KR (skramaja) wrote :

Simpler way to reproduce the problem in ansible http://paste.openstack.org/show/746470/

Revision history for this message
Saravanan KR (skramaja) wrote :

using "strategy: free" fixes the issue by ensuring the order of the execution. but it would increase the time of the execution, as execution will not be clubbed together for hosts having same var items.

default stragety (linear)
-------------------------
included: deployments.yaml for centos, localhost => (item=B)
included: deployments.yaml for centos, localhost => (item=C)
included: deployments.yaml for centos => (item=D)
included: deployments.yaml for localhost => (item=A)

with "strategy: free"
---------------------
included: deployments.yaml for centos => (item=B)
included: deployments.yaml for centos => (item=C)
included: deployments.yaml for centos => (item=D)
included: deployments.yaml for localhost => (item=A)
included: deployments.yaml for localhost => (item=B)
included: deployments.yaml for localhost => (item=C)

In the above example, the expected order for localhost is A, B, C.

Changed in tripleo:
milestone: stein-3 → train-1
Changed in tripleo:
milestone: train-1 → train-2
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix proposed to tripleo-heat-templates (master)

Fix proposed to branch: master
Review: https://review.opendev.org/665059

Changed in tripleo:
status: Triaged → In Progress
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Change abandoned on tripleo-heat-templates (master)

Change abandoned by Saravanan KR (<email address hidden>) on branch: master
Review: https://review.opendev.org/665059

Revision history for this message
Saravanan KR (skramaja) wrote :

This issue will be fixed as part of https://review.opendev.org/#/c/663710

Changed in tripleo:
status: In Progress → Fix Committed
wes hayutin (weshayutin)
Changed in tripleo:
status: Fix Committed → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.