This implementation should be more efficient because it only uses the
names under /proc, whereas the previous implementation fetched the
full directory entry for each one, and sorted them too.
New behaviour in Docker 1.12 means it disregards a plugin-supplied
address pool that overlaps an existing route, which causes a hang as
noted in #2469. When we pass in the subnet, Docker believes we really
do want that address range and avoids the hang.
The #2418 has changed a behaviour of "weave ps" which outputs
"Link not found" in the case of weave not running. The previous
behaviour was to output an empty line.
Make sure that any network namespace changes happens at the end of
the cmd invocation. Otherwise, we are risking to end up in a situation
in which OS scheduling threads are in a "wrong" namespace.
See WithNetNS for more details.
- Optimize. Use LinkByName instead of iterating over list of available
links. Linux does not support interfaces with the same names.
- Remove unnecessary indirection in utility functions which return
common.NetDev.
When running in the AWSVPC mode, we need to do the following:
- Keep the HW TX offloading on.
- Check that the allocated address belongs to the default subnet,
because only one subnet is allowed and it should be the default one.
If it is not possible to determine whether AWSVPC is on, we ignore
all the steps, because the attachment routines have to work
regardless whether the router is running and the AWSVPC detection
is done by querying the /ipinfo/tracker endpoint served by
the router.
Now it's off by default, and there is an option to enable the
multicast route per-network.
We turn it on for the `weave` network created when we start the plugin.
Note that we remove the ability to set the weavemesh network name, or
disabling it altogether. This was always poorly support, since the
removal operation only ever operated on the default name. The option
was also not listed in the usage, though it was in the docs.
Bryan Boreham