yahoo-eng-team team mailing list archive

[Launchpad Bug Tracker <1771662@xxxxxxxxxxxxxxxxxx>]

This bug was fixed in the package libvirt - 4.6.0-2ubuntu3.3

---------------
libvirt (4.6.0-2ubuntu3.3) cosmic; urgency=medium

  * d/p/ubuntu/lp-1811198-utils-Remove-arbitrary-limit-on-socket_id-core_id
    .patch: fix arm servers with high core_id (LP: #1811198)
  * d/p/ubuntu/lp-1771662-*: fix assumption that all VFs have PFs assigned
    (LP: #1771662)

 -- Christian Ehrhardt <christian.ehrhardt@xxxxxxxxxxxxx>  Thu, 31 Jan
2019 12:29:37 +0100

** Changed in: libvirt (Ubuntu Cosmic)
       Status: Fix Committed => Fix Released

** Changed in: libvirt (Ubuntu Bionic)
       Status: Fix Committed => Fix Released

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to OpenStack Compute (nova).
https://bugs.launchpad.net/bugs/1771662

Title:
  [bionic] libvirtError: Node device not found: no node device with
  matching name

Status in OpenStack nova-compute charm:
  Invalid
Status in OpenStack Compute (nova):
  Invalid
Status in libvirt package in Ubuntu:
  Fix Released
Status in libvirt source package in Bionic:
  Fix Released
Status in libvirt source package in Cosmic:
  Fix Released
Status in libvirt source package in Disco:
  Fix Released

Bug description:
  [Impact]

   * Libvirt has had the assumption that every VF (virtual function) will
     have a PF (physical function) assigned, but that does not hold true on
     some special Hardware like the Cavium ThunderX

   * Dannf helped some patches initially from Linaro to be accepted upstream
     and those we'd want to backport to Bionic and Cosmic

  [Test Case]

   * Verify that virsh nodedev-list shows the onboard NICs:
     $ sudo virsh nodedev-list | grep ^net
     net_enP2p1s0f1_42_ca_74_64_88_75
     net_enP2p1s0f2_42_ca_74_64_88_76
     net_enP2p1s0f3_42_ca_74_64_88_77

   * This needs plenty of setup and special HW, but Jason Hobbs & Dannf are
     willing to do the verification in our test lab.

  [Regression Potential]

   * Review hasn't spotted any issues, but in theory there could be negative
     effects to PF/VF pass-through cases. There is some code cleanup
     associated that should not, but might cause issues on that.
     I'd ask Jason to also run PF/VF workload on the PPA/SRU with other
     Hardware as well (like our x86 test environment) to be sure of that
     being ok.

  [Other Info]

   * n/a

  ---

  After deploying openstack on arm64 using bionic and queens, no
  hypervisors show upon. On my compute nodes, I have an error like:

  2018-05-16 19:23:08.165 282170 ERROR nova.compute.manager
  libvirtError: Node device not found: no node device with matching name
  'net_enP2p1s0f1_40_8d_5c_ba_b8_d2'

  In my /var/log/nova/nova-compute.log

  I'm not sure why this is happening - I don't use enP2p1s0f1 for
  anything.

  There are a lot of interesting messages about that interface in syslog:
  http://paste.ubuntu.com/p/8WT8NqCbCf/

  Here is my bundle: http://paste.ubuntu.com/p/fWWs6r8Nr5/

  The same bundle works fine for xenial-queens, with the source changed
  to the cloud-archive, and using stable charms rather than -next. I hit
  this same issue on bionic queens using either stable or next charms.

  This thread has some related info, I think:
  https://www.spinics.net/linux/fedora/libvir/msg160975.html

  This is with juju 2.4 beta 2.

  Package versions on affected system:
  http://paste.ubuntu.com/p/yfQH3KJzng/

To manage notifications about this bug go to:
https://bugs.launchpad.net/charm-nova-compute/+bug/1771662/+subscriptions