Fix missingImport logic to handle a module that cannot be imported, due
to a missing dependant module.
Due to the nature of this bug doctests cannot cover it.
Fix bug 1133103
Change-Id: Ia067bacc125b67e2d9e0cfc72495e4f13a8648ad
Change 1c5518ba introduced different user/passwd for the baremetal CI
tests, which did not match the CI systems. Since then, all the baremetal
migration tests have been skipped or failing.
This patch sets the user/passwd back to 'openstack_citest'.
Change-Id: I88508d83d6e9e63f20c26704ec6b455c48a33e8f
This patch adds support to linux_net and nova-dhcpbridge for
specifying and conveying multiple dhcpbridge_flagfile values.
The support is implemented as follows:
dhcpbridge_flagfile is now a "MultiStr" configuration file option
allowing the flag files to be defined over multiple configuration files.
The configured files are concatenated into a comma separated string and
assigned to the FLAGFILE environment variable when invoking
nova-dhcpbridge. nova-dhcpbridge splits the value of the environment
variable on the comma an initializes its configuration with the provided
files (the configuration system already supports loading configuration
from multiple files)
Fixes bug 1131932
Change-Id: Ic9cc57b8f7d9356606b9134ef5b63ab07b6099d9
Testr doesn't directly support running tests with pdb so the --debug
option was pulled from run_tests.sh during the testr switch. Add the
flag back into run_tests.sh and have it run tests with testtools.run
which allows users to import pdb in tests then run them without
interference from testr.
Example usage:
# First edit test file to import pdb.
run_tests.sh -d path.to.test
If the test is not given on the command line all tests will be run.
Change-Id: I4afca9ca997ccb6c575b714b778a9db8306c7891
Fixes bug #1132138
Only unused disk images are currently cleaned up by the image cache
manager but it seems logical to clean up unused kernels and ramdisks
too.
Achieve that by writing kernels and ramdisks to disk using the sha1
sum of their ID as the filename. This is the same scheme as used for
disk image filenames and causes the image cache manager to consider
them for cleanup. We also make the cache manager take note of in use
kernels and ramdisks when iterating over the list of instances.
A nasty upgrade concern is that if we immediately switch to writing
kernels to disk using this scheme then, where shared storage is used,
we can have older image cache managers on remote compute nodes cleaning
up kernels because they appear unused. To mitigate that, turn off this
behaviour by default and allow it to be enabled using a new config
option. This option will be removed in future and the behaviour enabled
by default.
DocImpact: new remove_unused_kernels option
Change-Id: I56bba9fa6596601104498e262c2e657f0eae2fa0
This patch uses to_primitive() to ensure that the macs argument to the
rpcapi method allocate_for_instance() can be serialized by rpc drivers.
Specifically, the powervm driver uses a set() here, and that will cause
breakage in non-kombu drivers.
Fix bug 1131035.
Change-Id: I5729c5854799e49c6598e4dd2a4da487598307f1
This patch adds a check to confirm that a security group name
is not among a list of 'special' group names before proceeding
with security group deletion.
Fixes: bug 1131830
Change-Id: I5656e01b4ed3fd9c22abe6a731edaef7203d97c4
lxml doesn't roundtrip empty strings, meaning if you serialize
and element with value '', when you derserialize it, the value will be
`None`.
This broke the version_string_with_package check because None and ''
will not compare as equivalent.
The solution is to coerce the None value back to an empty string before
running the check.
Change-Id: Ia3a8ae68f12a4645057d243fdee59197df40e8d3
Updates the OSAPI sizelimit middleware so that we use avoid calling
len on a request body which could cause a really large request
to get buffered into memory.
Also updates the middleware to return HTTP 413 which is a more
correct error code in this case (previously it returned just 400).
Fixes LP Bug #1131857.
Change-Id: Id8bc5eeb0fba9482809edd12543a75163e1227e9
This wrapper was there to allow oslo-config changes through the gate
while keystoneclient hadn't been updated.
Change-Id: I3d48673af58ae379e8e4d2c7fb4b4d81f3ed67dd
In patch ab9f8667c6, several
baremetal/db/api methods' exceptions were changed from InstanceNotFound
to NodeNotFound. The API extension for baremetal was not updated to
catch these, and this was not caught by unit testing. This resulted in
unhandled exceptions within nova-api if any baremetal node lacked an
associated interface.
While fixing that bug, a few other unit tests for the baremetal API
were added, and a missing exception was added to db/api bm_node_destroy.
Fixes bug 1131430.
Change-Id: I15f7624723754f9d7b217b609663a2d709acb056
Updates our libguestfs vfs implementation so that it makes use of
the same libvirt connection URI as Nova does when using
attach mode 'libvirt'. Previously when libguestfs used attach
mode 'libvirt' (the default on some newer distros like Fedora 18)
you could get permissions issues and ultimately get a libvirt
connection error from libguestfs:
could not connect to libvirt (URI = NULL)
Docs on guestfs attach method can be found here:
http://libguestfs.org/guestfs.3.html#attach-method
Having libguestfs use the same connection URL as Nova makes sense
and gets past these permission issues. (avoiding things like forcing
you to create a $HOME dir for the nova user among other things)
Full thread on this is here:
https://bugzilla.redhat.com/show_bug.cgi?id=913345
Note that on most distros the default libguestfs attach method
is still 'appliance' so this would have no affect unless explicitly
enabled via LIBGUESTFS_ATTACH_METHOD.
Fixes LP Bug #1131768.
Change-Id: I7e9f010022731a43e8d793cf8415ec12b346fdb4
LibvirtDriver's uri() function seems to be the most logical place
where one can access the libvirt URL used by Nova. I would like
access to this outside the scope of the internal libvirt
driver (for libguestfs) so exposing it as a static
resource seems to make sense.
Partial fix for LP Bug #1131768.
Change-Id: I653296414716b4e799d68f6bfc868e6039d6ea4a
When using the firewall driver IptablesFirewallDriver and the
default INPUT and FORWARD rules are DISCARD then the DHCP
request from the VM is discarded prior to getting to the dnsmasq.
A new rule will be added that enables DHCP requests to pass.
This fixes bug 1131223
Change-Id: I50fad5b63c3c4b22a5d828e3e89353c1ed723332
Fixes bug 1131033
Power VM driver is using qemu-img through the fetch_to_raw call
No need to attempt a conversion since power images can only
be represented with the raw format, removing thus the need for
any conversions by calling fetch instead of fetch_to_raw.
Change-Id: I94b46e184cb87924a4754a0a48d9c2013a3eb756
Jenkins