The file permissions can be specified (unfortunately) as a string or an
octal integer. During the normalization step, every field is
unmarshalled into an interface{}. String types are kept in tact but
integers are converted to decimal integers. If the raw config
represented the permissions as an octal, it would be converted to
decimal _before_ it was saved to RawFilePermissions. Permissions() would
then try to convert it again, assuming it was an octal. The new behavior
doesn't assume the radix of the number, allowing decimal and octal
input.
In all of the YAML tags, - has been replaced with _. normalizeConfig() and
normalizeKeys() have also been added to perform the normalization of the input
cloud-config.
As part of the normalization process, falsey values are converted to "false".
The "off" update strategy is no exception and as a result the "off" update
strategy has been changed to "false".
- Explicitly specify all of the valid options for etcd
- Remove the default name generation (ETCD_NAME is set by its unit file now)
- Seperate the etcd config from Units()
- Remove support for DISCOVERY_URL
- Add YAML tags for the fields
TestWriteEnvFileDos2Unix had a copy/paste bug, it shouldn't have
asserted that mtime doesn't change because the file is actually being
modified in this test. This didn't come up earlier because the actual
comparison wasn't using Time.Equal as it should have.
Instead switch to comparing inode numbers which is the actual thing I
wanted to test for in the first place, just accessing them is much more
awkard. Now all tests where it is relevant check the inode in addition
to the contents.
To maintain the behavior of the coreos-setup-environment that has
started to move into cloudinit we need to write out /etc/environment
with the public and private addresses, if known. The file is updated so
that other contents are not replaced. This behavior is disabled entirely
if /etc/environment was written by a write_files entry.
Until support for bonding params is added to networkd, this will be
neccessary in order to use bonding parameters (i.e. miimon, mode).
This also makes it such that the 8012q module will only be loaded if
the network config makes use of VLANs.
In order for networkd to properly configure the network interfaces, the configs must be
prefixed to ensure that they load in the correct order (parent interfaces have a lower
prefix than their children).
Adding the flag -convertNetconf which is used to specify the config
format to convert from (right now, only 'debian' is supported).
Once the network configs are generated, they are written to
systemd's runtime network directory and the network is restarted.
This change creates a few simple interfaces for coreos-specific
configuration options and moves things to them wherever possible; so if
an option needs to write a file, or create a unit, it is acted on
exactly the same way as every other file/unit that needs to be written
during the cloud configuration process.
Fix#69 - A user may provide an `enable` attribute of a unit in their
cloud config document. If true, coreos-cloudinit will instruct systemd
to enable the associated unit. If false, the unit will not be enabled.
Fix#71 - The default enable behavior has been changed from on to off.
When using openssh without pam it checks for a ! prefix in the password
field, locking the account entirely if found. The other common lock
character, *, is allowed by ssh to login via ssh keys so use it instead.
