| Age | Commit message (Collapse) | Author |
|---|
|
|
|
if you have UNCOMMITTED=1 in environment then the tree will
contain uncommitted changes.
|
|
Similar test-hack as the SmartOS one.
Here we just pretend that we're running on x86_64 so as to avoid
re-writing the tests to expect otherwise on arm.
|
|
|
|
As with SmartOS change earlier, running dmidecode on arm will crash kvm.
So instead of doing that, just return UNKNOWN which will cause this
data source to not activate.
LP: #1285686
|
|
See LP1243287 for more information
|
|
|
|
See LP: #1243287 for more information, but the easiest thing to do
here is just not run smartos on arm.
LP: #1243287
|
|
after freebsd merge, ubuntu shows addr:v.x.y.z instead of v.x.y.z for the
ipv4 address. This should fix that by just skipping the 'inet' (or inet6)
token if the next token starts with 'addr:'.
LP: #1285185
|
|
When cloud-init writes previous-instance-id, it does so with a
trailing '\n'. This isn't ideal, as other readers also have to know
that this will have a trailing '\n' on it, but here we just trim that off.
|
|
do not search 'latest', as that would be assuming the versioned
api would never break.
Also, instead of warning when we use something else, just debug.
|
|
|
|
The instance-id file contains the instance id
and a newline, to compare correctly make sure
we strip the newline before further usage.
|
|
This test case isn't really valid any more as we've changed
the way this works.
|
|
|
|
The CloudSigma datasource would attempt to read /dev/ttyS1
and if not found would warn (which gets logged to stdout by default).
Better to just debug.
|
|
|
|
There might be multiple things to put inside a vendor-data.
So, if it is a dict and that dict has 'cloud-init', assume that the whole
thing was not meant for cloud-init, and set vendordata_raw to the specific
item.
|
|
The code in the previous commit was creating /var/lib/cloud/instance/
when it should not have. This is better handled now by using
/var/lib/cloud/instances/<instance-id>, and then letting the link get
created by cloud-init elsewhere.
|
|
datasource for SmartOS runs. This patch creates the instance's directory.
|
|
This reduces how much cloud-init is explicitly involved in what "vendor-data"
could accomplish. The goal of vendor-data was to provide the vendor with a
channel to run arbitrary code that accomodate for their specific platform.
Much of those accomodations are currently being done in cloud-init.
However, this now moves some of those things to default "vendor-data", instead
of cloud-init proper.
Basically, now we have an 'sdc:vendor-data' key in the metadata.
If that does not exist, then cloud-init will use the default.
The default, provides a boothook. That boothook writes a file into
/var/lib/cloud/per-boot/ . That file will be both written on every boot
and then executed at rc.local time frame (by 'scripts-per-boot').
It will then execute /var/lib/cloud/instance/data/user-script
and /var/lib/cloud/instance/data/operator-script if they exist.
So, the things that cloud-init is now doing outside of the default vendor-data
that I would rather be done in vendor-data is:
* managing the population of instance/data/user-script and
instance/data/operator-script. These could very easily be done
from the boothook, but doing them in cloud-init removes the necessity
for having a 'mdata-get' command in the image (or some other way for
the boothook script to query the datasource).
* managing the LEGACY things.
|
|
This reduces how much cloud-init is explicitly involved in what "vendor-data"
could accomplish. The goal of vendor-data was to provide the vendor with a
channel to run arbitrary code that accomodate for their specific platform.
Much of those accomodations are currently being done in cloud-init.
However, this now moves some of those things to default "vendor-data", instead
of cloud-init proper.
Basically, now we have an 'sdc:vendor-data' key in the metadata.
If that does not exist, then cloud-init will use the default.
The default, provides a boothook. That boothook writes a file into
/var/lib/cloud/per-boot/ . That file will be both written on every boot
and then executed at rc.local time frame (by 'scripts-per-boot').
It will then execute /var/lib/cloud/instance/data/user-script
and /var/lib/cloud/instance/data/operator-script if they exist.
So, the things that cloud-init is now doing outside of the default vendor-data
that I would rather be done in vendor-data is:
* managing the population of instance/data/user-script and
instance/data/operator-script. These could very easily be done
from the boothook, but doing them in cloud-init removes the necessity
for having a 'mdata-get' command in the image (or some other way for
the boothook script to query the datasource).
* managing the LEGACY things.
|
|
|
|
|
|
this changes url_map to a list and adds 'required' information.
* If we've not already found an entry, and this is required,
then debug log (ie, this is just not GCE).
* if we already found an entry and this is required: warn
split the keys fixing out of the loop.
|
|
default #cloud-config that writes per-boot script that fetches subsequent
sdc:operator-scripts and executes it.
|
|
|
|
|
|
Openstack has a unique derivative datasource that is gaining usage.
Previously the config drive datasource provided part of this functionality
as well as the ec2 datasource, but since new functionality is being added
to openstack's special datasource it seems beneficial to combine the used
parts into a new datasource just made for handling openstack deployments
that use the openstack metadata service (possibly in combination with the
ec2 metadata service).
This patch factors out the common logic shared between the config drive
and the openstack metadata datasource and places that in a shared helper
file and then creates a new openstack datasource that readers from the
openstack metadata service and refactors the config drive datasource to
use this common logic.
|
|
|
|
|
|
Waiting around for a metadata service in a given datasource means that if its
not there all the subsequent datasources have to wait, and boot is slowed down.
As it is right now, EC2 is the only one that has the right to wait. In the
past, we had to wait around for the EC2 metadata service. I really do not want
to extend that courtesy to other cloud platforms. A network based metadata
service should be up as soon as networking is up.
|
|
|
|
|
|
There are some rough edges here and its missing some test, but
I want to get this pulled in.
|
|
This just adds user-data in 'instance/attributes/user-data'.
Also turns retries to 0 on all other things.
|
|
Previously this had 'local' as the default datasource mode, meaning
that user-data code such as boot hooks and such would not be guaranteed to have
network access. That would be out of sync with the expectation on other
platforms where the default is 'network up'.
The user can still specify 'dsmode' as local if necessary and the
local datasource will claim itself found.
|
|
|
|
this allows the metadata url to be
configured by setting:
datasource:
GCE:
metadata_url: <value>
Then also, if its not resolvable, we just deactivate the datasource quickly.
|
|
|
|
|
|
|
|
|
|
Config modules are able to declare distros that they were verified
to run on by setting 'distros' as a list in the config module.
Previously, if a module was configured to run and the running distro was not
listed as supported, it would run anyway, and a warning would be written.
Now, we change the behavior to skip those modules.
The distro (or user) can specify that a given list of modules should run anyway
by declaring the 'unverified_modules' config variable.
run_once modules will be run without this filter (ie, expecting that the user
explicitly wanted to run it).
|
|
for now, this the mechanism just doesn't seem right.
I think i'd rather have the module declare supported distros than
have distros declare [un]supported modules.
|
|
|
|
this module should "work" everywhere, in that it will only do anything if
/sbin/initctl exists (which is going to be upstart).
|
|
|
|
|
|
LP: #1277746
|
