| Age | Commit message (Collapse) | Author |
|---|
|
|
|
The one issue i'm aware of currently is that tap devices
(ip tuntap add mode tap user root mytap1)
do not work correctly with 'is_up' which means the check
does not bring them down and the rename fails.
The LOG.debug message should be cleaned up too, as it currently
references the function rather function.__name__ for nicer message.
|
|
currently does not work in lxc
https://github.com/lxc/lxd/issues/2063
|
|
== background ==
DataSource Mode (dsmode) is present in many datasources in cloud-init.
dsmode was originally added to cloud-init to specify when this datasource
should be 'realized'.
cloud-init has 4 stages of boot.
a.) cloud-init --local . network is guaranteed not present.
b.) cloud-init (--network). network is guaranteed present.
c.) cloud-config
d.) cloud-init final
'init_modules' [1] are run "as early as possible". And as such, are executed
in either 'a' or 'b' based on the datasource. However, executing them means
that user-data has been fully consumed. User-data and vendor-data may have
'#include http://...' which then rely on the network being present. boothooks
are an example of the things run in init_modules.
The 'dsmode' was a way for a user to indicate that init_modules
should run at 'a' (dsmode=local) or 'b' (dsmode=net) directly.
Things were further confused when a datasource could provide networking
configuration. Then, we needed to apply the networking config at 'a'
but if the user had provided boothooks that expected networking, then the
init_modules would need to be executed at 'b'. The config drive datasource
hacked its way through this and applies networking if *it* detects it is
a new instance.
== Suggested Change ==
The plan is to
1. incorporate 'dsmode' into DataSource superclass
2. make all existing datasources default to network
3. apply any networking configuration from a datasource on first boot only
apply_networking will always rename network devices when it runs.
for bug 1579130.
4. run init_modules at cloud-init (network) time frame unless datasource
is 'local'.
5. Datasources can provide a 'first_boot' method that will be called when
a new instance_id is found. This will allow the config drive's write_files
to be applied once.
Over all, this will very much simplify things. We'll no longer have
2 sources like DataSourceNoCloud and DataSourceNoCloudNet, but would just
have one source with a dsmode.
== Concerns ==
Some things have odd reliance on dsmode. For example, OpenNebula's get_hostname
uses it to determine if it should do a lookup of an ip address.
== Bugs to fix here ==
http://pad.lv/1577982 ConfigDrive: cloud-init fails to configure network from network_data.json
http://pad.lv/1579130 need to support systemd.link renaming of devices in container
http://pad.lv/1577844 Drop unnecessary blocking of all net udev rules
|
|
|
|
When ip= on the kernel command line defines the networking, set
those network devices to be manually controlled, instead of 'auto'.
The reason for this is that if they're marked as 'auto':
a.) a second attempt will be made to ifup them.
b.) they'll be brought down on shutdown
'b' is problematic on network root filesystem.
Also this picks up 2 changes from curtin's net module:
- Cleanup newline logic so we always have a clean '\n\n' between stanza
- Add a unittest to validate bonding network config render, specifically
when to emit auto $iface for dependent bond slaves.
LP: #1568637
|
|
|
|
when reading the initramfs configurewd devices and turning them
into network config, we change to not have 'auto' control (or allow=auto).
The reason for this is that if the device was still up:
a.) it would try to bring it up again (due to bug 1570142)
b.) it would be brought down.
'b' is problematic if there is an iscsi or network root filesystem.
Note, that ifupdown does now support 'no-auto-down' which means
that the nic should not be brought down on 'ifdown -a'.
LP: #1568637
|
|
This picks up newline cleanup and some bond fixes from curtin at rev 374.
- Cleanup newline logic so we always have a clean '\n\n' between stanza
- Add a unittest to validate bonding network config render, specifically
when to emit auto $iface for dependent bond slaves.
|
|
Just skip devices that are named veth*.
The fix here is to ignore lxd created devices, but any other veth
device that is created at this point in boot is probably not the
right interface to dhcp on.
LP: #1569064
|
|
It does not make sense to consider bridges when searching for fallback
networking. If the system is configured with a bridge, then its probably
for some purpose other than to get to a metadata service.
Considering the bridge could make cloud-init pick the wrong device on reboot.
LP: #1569974
|
|
a.) do not write systemd link files if we do not have a mac address.
the check is updated to check for value rather than just presense
(ie, 'mac_address': None)
b.) DataSourceNoCloudNet: search in the nocloud seed dir
this is important because NoCloud if dsmode is Net will look only
would pass by, expecting NoCloudNet to pick it up
but NoCloudNet would not look in /var/lib/cloud/seed/nocloud
and thus skip it.
c.) support the disabling of network configuration
via /var/lib/cloud/data/upgraded-network
This is what the package upgrader is writing.
|
|
|
|
This adds support for suppling network configuration on the
kernel command line in 2 ways:
a.) kernel command line includes 'network-config=<base64>'
value of that parameter is base64 encoded json (or yaml)
it is taken as network config yaml.
In order to save space on kernel command line, it can be
base64 encoded gzipped json also.
b.) ip= paired with files authored by klibc's ipconfig tool
When network devices are brought up in the initramfs, klibc's
ipconfig tool writes files are named /run/net-<DEVNAME>.conf.
The best documentation available on that tool is
/usr/share/doc/libklibc/README.ipconfig.gz.
Also changes util.get_cmdline() to return the command line of
pid 1 if it is in a container. That is to make it consistent with
The systemd generator, and allow passing a command line to lxd,
as lxd does not mask /proc/cmdline.
|
|
|
|
net: add render_route comment to document why we added || true to route
statements
DataSourceConfigDrive: Only convert network_json to network_config when
caller reads network_config attr. Cache the conversion.
|
|
|
|
add tests to show this functional.
|
|
|
|
|
|
|
|
|
|
Wesley's loader returned network state, so that got me updating
it, and i implemented as such. Then realized that actually
ipconfig (klibc) has no support for ipv6. So even though
i painfully generalized that, its pointless. next commit will
drop it.
|
|
|
|
|
|
|
|
Tested now with the generated fallback config in an lxc container.
Had to change to return a config rather than a network state.
Also this makes nocloud look in nocloud-net's seed dir.
This way it will read the seed and clame the datasource but
not do anything other than apply networking and the init_modules early.
It is a change in behavior of the time that boothooks woudl run to do
this. May need to change that back.
|
|
|
|
there is no data source that has a populated network_config()
so at this point this doesn't do anything.
|
|
on interfaces with characters between their numbers
|
|
'version' keys
|
|
- Removed systemd .link file generation, as it is not needed right now
- Changed return of generate_fallback_config to be just ns dict
- In distros.debian don't attempt to write .link file
|
|
|
|
not to attempt renaming the network interface to the default interface. Default
interface is controleld by net.DEFAULT_PRIMARY_INTERFACE and is currently set
to eth0 for legacy reasons. By default cloud-init will not attempt to rename
the device as this does not work in some situtations depending on the backing
driver of the device.
|
|
|
|
try/except blocks because there are sometimes read errors on the files and this
should not cause a stacktrace
|
|
on in the event that no network configuration was provided to cloud-init
- Devices in /sys/class/net aside from loopback devices are scanned
- Each device is tested to determine if it has a carrier using
/sys/class/net/DEV/carrier, devices which do are preferred as they are most
likely connected to the outside world
- Devices which do not have a carrier but which might still be connected due
to being in a dormant or down state are used as fallbacks in case no
devices are found which have a carrier
- A network state dictionary is generated to be passed to
render_network_state to write ENI
- A systemd link file is generated that will rename the chosen device to eth0
|
|
|
|
|
|
lp:cloud-initramfs-tools/dyn-netconf/scripts/init-bottom/cloud-initramfs-dyn-netconf
|
|
broadcast, netmask, gateway and hostname if present
|
|
kernel's cmdline during network configuration parsing.
- Search for .conf files in /run with names starting with 'net', as these are
created during early boot if the ip parameter is present
- If any are present and valid they are merged with network configuration
from the current data source
- If the devices affected by the 'ip' parameter are already present in
network configuration, then a subnet entry will be added to the device's
configuration unless an identical entry is already present
- If any of the devices affected are not present then a mostly blank
configuration will be generated for the device and the appropriate subnet
specified
|
|
at this point, this works:
python -m cloudinit.net.network_state examples/network-all.yaml
|
|
just add curtin/net as cloudinit/net
and then copy curtin/udev.py as cloudinit/net/udev.py
|
