libvirt driver doesn't cleanup the tap interface on vm re-schedule

[pritesh]

Here when you use libvirt driver with tap interfaces, it creates a tap interface on the host but doesn't clean up the interface and leaves in-tact and creates another same named interface on the new host.

In _do_build_and_run_instance when RescheduledException is called, manager checks if the network port needs to be de-allocated for a different host or not using deallocate_networks_on_reschedule() which is hard coded to return False. If this is changed to return true or set via conf file configuration to allow being changed for specific mech drivers in neutron then it would be helpful to not only clean up the tap interface properly but also also mech drivers in neutron to re-create new ports on new host instead of shifting and re-using same ports which fails.

tested on master and stable/liberty and fails in both cases, so may need back porting.

[Wenzhi Yu (yuywz)]

Duplicated with this bug: https://bugs.launchpad.net/nova/+bug/1510979

[pritesh (pritesh)]

@Wen: this is not a duplicate bug, I tested the patch (https://review.openstack.org/243477) which https://bugs.launchpad.net/nova/+bug/1510979 fixes, but it does not cover a fix for tap interfaces, it just fixes it for neutron. I have attached log files to show the latest test. the tap interface are created with same name on multiple hosts like shown here:

Host-1:
ubuntu@ip-192-168-0-10:~$ ifconfig tapcaba1b94-47
tapcaba1b94-47 Link encap:Ethernet HWaddr c6:be:57:50:ba:93
          inet6 addr: fe80::c4be:57ff:fe50:ba93/64 Scope:Link
          UP BROADCAST MULTICAST MTU:1500 Metric:1
          RX packets:0 errors:0 dropped:0 overruns:0 frame:0
          TX packets:0 errors:0 dropped:2 overruns:0 carrier:0
          collisions:0 txqueuelen:500
          RX bytes:0 (0.0 B) TX bytes:0 (0.0 B)

Host-2:
ubuntu@ip-192-168-0-11:~$ ifconfig tapcaba1b94-47
tapcaba1b94-47 Link encap:Ethernet HWaddr a6:07:7c:bc:ce:e7
          inet6 addr: fe80::a407:7cff:febc:cee7/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
          RX packets:9 errors:0 dropped:0 overruns:0 frame:0
          TX packets:8 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:500
          RX bytes:1467 (1.4 KB) TX bytes:648 (648.0 B)

[Andrew Laski (alaski)]

There's been no progress in 4 months. Sana if you are still working on this please assign it back to yourself.

[Sujitha (sujitha-neti)]

Sana Khan, If you are currently working on a patch, can you please change the status of this bug to In Progress.

[Augustina Ragwitz (auggy)]

Also be sure to use the tag "Closes-Bug:#1531022" in your commit message so the bug is automatically updated.

[Eric Fried (efried)]

Addressed in nova-powervm by https://review.openstack.org/#/c/460672/

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/ocata
Review: https://review.openstack.org/461653

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/460672
Committed: https://git.openstack.org/cgit/openstack/nova-powervm/commit/?id=dedc355805c24583fbad5f4e5c416881239e377e
Submitter: Jenkins
Branch: master

commit dedc355805c24583fbad5f4e5c416881239e377e
Author: <email address hidden> <email address hidden>
Date: Thu Apr 27 12:43:07 2017 -0400

    Deallocate network on reschedule

    When a virtual machine fails to deploy on one host
    then deallocate networks so that when the same
    virtual machine deployed on another host network
    is allocated again.

    Change-Id: I9624e44def8afa01c45527d625917adb6847c55f
    Closes-bug: #1531022

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/461653
Committed: https://git.openstack.org/cgit/openstack/nova-powervm/commit/?id=40df0771efff1f8c4890d2c1c63e23801629b465
Submitter: Jenkins
Branch: stable/ocata

commit 40df0771efff1f8c4890d2c1c63e23801629b465
Author: <email address hidden> <email address hidden>
Date: Thu Apr 27 12:43:07 2017 -0400

    Deallocate network on reschedule

    When a virtual machine fails to deploy on one host
    then deallocate networks so that when the same
    virtual machine deployed on another host network
    is allocated again.

    Change-Id: I9624e44def8afa01c45527d625917adb6847c55f
    Closes-bug: #1531022

[Sean Dague (sdague)]

There are no currently open reviews on this bug, changing
the status back to the previous state and unassigning. If
there are active reviews related to this bug, please include
links in comments.

[Sean Dague (sdague)]

Automatically discovered version liberty in description. If this is incorrect, please update the description to include 'nova version: ...'

[Matt Riedemann (mriedem)]

I'm going to remove 'nova' from this since we can track the nova issue with bug 1597596. The powervm driver out of tree has fixed this independently.