yahoo-eng-team team mailing list archive
-
yahoo-eng-team team
-
Mailing list archive
-
Message #55975
[Bug 1616782] Re: driver_discard='unmap' is written without considering the configuration
This is not a bug - there are two completely independent settings here
- nova.conf hw_disk_discard is applicable to compute node local ephemeral storage
- cinder 'discard' property is applicable to cinder volume storage
Using hw_disk_discard when configuring cinder volumes is not
appropropriate.
** Changed in: nova
Status: In Progress => 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/1616782
Title:
driver_discard='unmap' is written without considering the
configuration
Status in OpenStack Compute (nova):
Invalid
Bug description:
The original code is as follows:
if data.get('discard', False) is True:
min_qemu = nova.virt.libvirt.driver.MIN_QEMU_DISCARD_VERSION
if self.connection._host.has_min_version(
hv_ver=min_qemu,
hv_type=host.HV_DRIVER_QEMU):
conf.driver_discard = 'unmap'
else:
global SHOULD_LOG_DISCARD_WARNING
if SHOULD_LOG_DISCARD_WARNING:
SHOULD_LOG_DISCARD_WARNING = False
LOG.warning(_LW('Unable to attach %(type)s volume '
'%(serial)s with discard enabled: qemu '
'%(qemu)s or later is required.'),
{
'qemu': min_qemu,
'serial': conf.serial,
'type': connection_info['driver_volume_type']
})
https://github.com/openstack/nova/blob/master/nova/virt/libvirt/volume/volume.py#L96
No consideration is given to the case when the configuration file is
not configured as hw_disk_discard=unmap
Only when the backend report can support discard, is to write the
unmap in XML
To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1616782/+subscriptions
References