Subnets in fullstack tests collide and collisions are resolved manually.

Bug #1562966 reported by Omer Anson
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
DragonFlow
Fix Released
Wishlist
Hong Hui Xiao

Bug Description

Subnets used in fullstack tests have overlapping IP ranges. These collisions are resolved manually.

This bug is to recommend using Neutron subnet-pools to allocate non-overlapping subnets in an automatic, extendible, and straight-forward manner.
For subnet pools, see:

http://blog.siliconloons.com/posts/2015-04-28-subnetpools/
http://specs.openstack.org/openstack/neutron-specs/specs/kilo/subnet-allocation.html
http://blog.episodicgenius.com/post/neutron-subnet-pools/
http://developer.openstack.org/api-ref-networking-v2-ext.html

Yuli (stremovsky)
Changed in dragonflow:
assignee: nobody → Pavel Gluschak (scsnow)
Revision history for this message
Pavel Gluschak (scsnow) wrote :

I've just started looking into this. I'm wondering how overlapping IP ranges in subnets affect fullstack tests execution?
I see, that tests are executed sequentially and most of tests do clean up of created subnet (if any) at the end.

Revision history for this message
Omer Anson (omer-anson) wrote :

fullstack tests run serially, now. But this was not always the case, and it works faster when the tests are run in parallel. To be able to support this scenario, we want subnets not to overlap.

Additionally, cleanup does not always work.

Revision history for this message
Yuli (stremovsky) wrote :

Hello

I think this task is a good idea.

The current hard coded value are not perfect but help us to detect a failing test very quickly.

Bu just checking object that are not deleted, you can easily detect what tests had fail.

Best regards,
Yuli

Revision history for this message
Yuli (stremovsky) wrote :

Ops, the opposite:

I do not think that this task is a good idea :)

Revision history for this message
Pavel Gluschak (scsnow) wrote :

I agree, that allocated resources should not be deleted if test fails. However these resources may affect subsequent tests.
But successful testcase should not leave any created resources. If that's the case, it should be fixed.

Pavel Gluschak (scsnow)
Changed in dragonflow:
assignee: Pavel Gluschak (scsnow) → nobody
Revision history for this message
Hong Hui Xiao (xiaohhui) wrote :

I think default subnetpool can be used. And I think identifing problem should use log, not some hard code string in the code.

https://review.openstack.org/#/c/368375/

Changed in dragonflow:
assignee: nobody → Hong Hui Xiao (xiaohhui)
status: New → In Progress
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix merged to dragonflow (master)

Reviewed: https://review.openstack.org/368375
Committed: https://git.openstack.org/cgit/openstack/dragonflow/commit/?id=2fe96282e3e934f3495ad8b301748d879420abb8
Submitter: Jenkins
Branch: master

commit 2fe96282e3e934f3495ad8b301748d879420abb8
Author: Hong Hui Xiao <email address hidden>
Date: Sun Sep 11 09:43:47 2016 +0800

    Add default-subnetpools to DFPlugin

    default subnetpools is pure control plane functionality in neutron. df can
    just enable it. And fullstack can use it to avoid subnet collisions.

    TODO: should we remove all unintentional, explicit subnet cidr defination
    in fullstack?

    Change-Id: I32afc73c1cf289bdbfb51ea3c140aff903c7261b
    Closes-Bug: #1562966

Changed in dragonflow:
status: In Progress → 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.