yahoo-eng-team team mailing list archive

[Juliane Aulbach <juliane.aulbach@xxxxxxxxxx>]

Bug is invalid, because it only occurs when Openstack is bypassed.
Vladiks solution is a great option for everybody who wants to prevent that. 

** Changed in: nova
       Status: Confirmed => Invalid

-- 
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/1443501

Title:
  Ejecting virtio-scsi attached devices causes inconsistent states

Status in OpenStack Compute (nova):
  Invalid

Bug description:
  Issue
  =====

  When ejecting virtio-scsi attached devices like the config-drive 
  the mount breaks. 

  The mount point delivers the following information:
  [27289.999820] sr 3:0:0:4: [sr0] Device not ready
  [27289.999832] sr 3:0:0:4: [sr0]  
  [27289.999834] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
  [27289.999836] sr 3:0:0:4: [sr0]  
  [27289.999838] Sense Key : Not Ready [current] 
  [27289.999840] sr 3:0:0:4: [sr0]  
  [27289.999844] Add. Sense: Medium not present
  [27289.999846] sr 3:0:0:4: [sr0] CDB: 
  [27289.999847] Read(10): 28 00 00 00 00 1c 00 00 01 00
  [27289.999853] end_request: I/O error, dev sr0, sector 112

  Inserting the device again is sometimes not possible.

  
  Steps to reproduce
  ==================

  * Create instance with config drive attached on a z node
  * Mount config drive (sr0) 
  * Eject config drive from the host
     virsh change-media <domain id> /opt/stack/data/nova/instances/<instance id>/disk.config --eject
  * Take a look at the mount point

  Expected behavior
  =================

  Device ejected. Mount point empty. Inserting config drive again is
  possible.

  Actual behavior
  ===============

  Two ejects necessary to eject device. Mount point breaks. Inserting
  config drive again is sometimes not possible. Sometimes it needs two
  inserts.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1443501/+subscriptions