This commit is a workaround due to how older debian and ubuntu systems
fail to properly restart networking. Instead of relying on the init
scripts or ifup/down tools to restart each interface, this commit
instead restarts each interface individually
This commit adds some additional logic that falls back to using the
ifdown/ifup tools to restart networking. On Ubuntu 14.04, the init
script was designed to always fail to restart newtorking, so it needs
to use the ifdown/up tools instead. This commit will use the networking
init script as a last resort to restart networking, assuming other
commands haven't broken networking already.
https://bugs.launchpad.net/ubuntu/+source/ifupdown/+bug/1301015
Prior to this commit, the hostname was set with one big bash script and
attempted to determine what tools are available. This commit changes
that by splitting out that tool check on the Vagrant side of things with
the GuestInspection class, and adds back restarting networking to get a
DHCP lease with the change rather than using `dhclient`. This pattern
matches how hostnames are set in the redhat capability.
If you have a vagrant box with proc mounted with
proc /proc proc defaults,hidepid=2 0 0
ps output will be limited to owned process
sudo should extend output
`vagrant up` may hang at the "Configuring and enabling network
interfaces..." step when private networks and PTY allocation for SSH
are used.
The newer version of `nmcli` that is part of CentOS now will open a
pager (i.e. `less`) for certain commands if it finds a tty. This
causes the invocations of `nmcli` in `guest_inspection.rb` to hang.
`nmcli` disables the use of a pager in 'terse' (`-t`) output mode,
while still returning enough information for the uses in
`guest_inspection.rb`.
Prior to this commit, when setting up private networks on Ubuntu using
netplan, it assumed that the guest was using systemd, the suggested
default tool to manage networking, and did not take into account devices
that could be managed with NetworkManager. This commit fixes that by
looking at the devices managed on the guest to see if its managed by
NetworkManager, and if so, use that renderer for netplan instead of
networkd.
Prior to this commit, the regex matcher used with grep to determine if a
system was using systemd? was failing on systems that did not exactly
match the old regex. This commit updates that communications test to use
a different method of determining if systemd is in use with the ps
command.
This updates the guest capabilities to run in as few communicator
commands as possible. Additionally, it fixes a number of issues around
hostname and idempotency.
This patch was tested against:
- puphpet/debian75-x64
- debian/jessie64
- debian/wheezy64
with custom networking, custom hostname, and rsync shared folders.
This fixes a fairly large tempfile leak. Vagrant uses a template
renderer to write network configuration files locally to disk. Then,
that temporarily file is uploaded to the remote host and moved into
place. Since Vagrant is such a short-lived process, GC never came along
and cleaned up those tempfiles, resulting in many temporary files being
created through regular Vagrant usage.
The Util::Tempfile class uses a block to ensure the temporary file is
deleted when the block finishes. This API required small tweaks to the
usage, but provides more safety to ensure the files are deleted.
When `/etc/hosts` contained a FQDN with a trailing dot, the `\b` in the
sed expression would not match, since dot is not considered to be a word
character.
Fix this by regexp-escaping the hostname search, and matching the end of
the line on optional space followed by additional characters.
Also add some tests that extract the regexp used by sed and verify that
it does what we want. These will hopefully serve us in the future if we
ever need to test additional edge cases.
there's been a lot of churn around this code, so i figure it was worth
trying to clean it up.
- the methods were doing a lot, so make them into template methods with
one helper per step
- spread out /etc/hosts regexp into a couple of helper variables for
clarity
- remove handling for broken hostname implementations (like basing all
of the checks on name.split('.')[0]), since it seems reasonable to
remove code dedicated only to handling broken boxes
- DRY up the shared code between debian/ubuntu implementations, which
clarifies the differences as well
- add unit tests around the behavior; this will help us in the future
to separate flaws in our understanding from flaws in implementation
- includes a new DummyCommunicator in tests which should be useful in
supporting additional unit testing of this kind
- manually tested this on squeeze, wheezy, precise, quantal, raring,
and saucy successfully.
handles the issue in #2333
Brian Cain
Chris Roberts
Maxim Kostrikin
Chris Roberts
Simon Baatz
Trey Tabner
Seth Vargo
Fabio Rehm
Paul Hinze