An environment with AF_INET6 sockets but no loopback interface creates
nothing but pain.
If an IPv4 server binds to 0.0.0.0:8080, clients may connect to
0.0.0.0:8080, which automatically picks 127.0.0.1 as a source address.
However, when a server binds to [::]:8080, the absence of ::1 causes
clients to fail with ENETUNREACH.
For a demonstration, run the following in a python shell:
import socket
s = socket.socket(socket.AF_INET6, socket.SOCK_STREAM, 0)
s.bind(("", 0))
print s.getsockname() # Example: ('::', 39079, 0, 0)
s.listen(10)
c = socket.socket(socket.AF_INET6, socket.SOCK_STREAM, 0)
c.connect(s.getsockname())
print c.getsockname(), c.getpeername()
This yields the following error:
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/usr/lib/python2.7/socket.py", line 224, in meth
return getattr(self._sock,name)(*args)
socket.error: [Errno 101] Network is unreachable
This DHCP exit hook to shorten the system hostname on GCE was previously
installed by build-debian-cloud and bootstrap-vz, but seems to have been
inadvertently removed in commit c81045cc6e
as part of a broad cross-cloud cleanup. Again, this was caught by our
validation tests, and we might be done with the fixes at this point.
In this commit, I'm reinstating the hook with a name change and an
explanatory comment, to reduce the risk of this vanishing accidentally
in the future.
Change-Id: I4e7268f8b9ab3b2a7fc8b510898c6fbdd685aa53
* force kpartx to use synchronous mode, removing sleep(10)
* get image configuration, use it during tarball creation and registration
* add (non-working, path problems) image registration
* add cleaning of image from Google keys and repositories
* add NTP server address in manifest
* add preference for backport kernels in manifest
* disable IPv6
* correctly set host name
