well. Add in additional GCE guest package into manifests (init files
were split into their own package to make the python package adhere to
pypi standards).
The way boot options for linux and config params for grub
were configured gave rise to quite a few bugs.
The configuration has now been abstracted so that
options can be added without interfering with the work
of other tasks (no more sed_i!)
Up until now I didn't see the point of using spaces for indentation.
However, the previous commit (a18bec3) was quite eye opening.
Given that python is an indentation aware language, the amount of
mistakes that went unnoticed because tabs and spaces were used
at the same time (tabs for indentation and spaces for alignment)
were unacceptable.
E101,W191 have been re-enable in the tox flake8 checker and
the documentation has been modified accordingly.
The following files have been left as-is:
* bootstrapvz/common/assets/extlinux/extlinux.conf
* bootstrapvz/common/assets/init.d/expand-root
* bootstrapvz/common/assets/init.d/generate-ssh-hostkeys
* bootstrapvz/common/assets/init.d/squeeze/generate-ssh-hostkeys
* bootstrapvz/plugins/docker_daemon/assets/init.d/docker
* bootstrapvz/providers/ec2/assets/bin/growpart
* bootstrapvz/providers/ec2/assets/grub.d/40_custom
* bootstrapvz/providers/ec2/assets/init.d/ec2-get-credentials
* bootstrapvz/providers/ec2/assets/init.d/ec2-run-user-data
* docs/_static/taskoverview.coffee
* docs/_static/taskoverview.less
* tests/unit/subprocess.sh
We were hard coding an image name format for GCE, but that doesn't match
what other providers (e.g. AWS) do.
Given that image upload did not work at all previously, no reason to
preserve the existing unusual behaviour.
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
All provider specific settings have been moved to the
provider section. The image name itself is now located
at the top level and called "name". It is required for all providers.
This change makes two fixes. First, it fixes a race condition on Debian
8 GCE instances which caused instances with >2TB disks to fail. Second,
it adds the required "PREREQ" header to the script, which was previously
missing and caused errors at build time. These two issues are somewhat
unrelated to each other, but debugging the first led to finding the
second.
