There's no Loopback CNI plugin for Windows as far as I can see, so I'm
assuming it's not needed.
Signed-off-by: Paul "TBBle" Hampson <Paul.Hampson@Pobox.com>
The `WithLoNetwork` and `WithConfFile` options consume the
`WithPluginDir` and `WithInterfacePrefix` settings, so they must come
first.
This fixes the issue that the provided plugin directory was not used for
the provided configuration file. This would have been more visible, but
the default `InterfacePrefix` is already "eth".
Signed-off-by: Paul "TBBle" Hampson <Paul.Hampson@Pobox.com>
The Windows API is completely different, so the common code cannot
create the unique identifier, but must accept what the OS-specific code
does.
For Linux, it's a pathname. In Windows, it's a GUID string.
This pulls in direct usage of Microsoft/hcsshim, which previously was
only an indirect usage via containerd calls from, e.g., localMounter on
Windows.
Signed-off-by: Paul "TBBle" Hampson <Paul.Hampson@Pobox.com>
Before this change, walkAllResults did not skip an item if it had already been
visited, so walking the graph of results actually followed every single possible
path in the graph instead of just visiting each item once.
On small graphs this wasn't noticeable, but on sufficiently large remote cache
imports this rapidly scaled out of control. For example, I first encountered
this when importing a max-cache from a build of a full linux rootfs from source
(which takes about 30 minutes to build w/out cache on an 18-core machine) and
after 30 minutes the cache import was still running with all CPUs pegged at 100%
To fix this, walkAllResults now keeps a map of already visited items (keyed by
their pointer) and skips visiting an item if it's already been visited, making
it usable on large remote cache imports.
Signed-off-by: Erik Sipsma <erik@sipsma.dev>
A recent change updated the containerd shim built for integ tests to the newer
v2, but the integ test code itself was still hardcoded to look for the older
shim binary and, when that binary wasn't found, skip running the test. This
caused all containerd tests to be skipped.
The fix here updates the binary name to v2 and also removes the branch that
will skip a test when the requirements check fails.
Signed-off-by: Erik Sipsma <erik@sipsma.dev>
The CI script goes looking for the branch to which a commit was pushed,
but was applying the branch name to moby/buildkit even if being built on
a different fork.
Signed-off-by: Paul "TBBle" Hampson <Paul.Hampson@Pobox.com>
Due to current deficiencies in the Windows implementation of
containerd's Mount.Mount, we need different behaviour for mounting on
Windows and Unix platforms.
Specifically:
* Windows mounts can only mount in-place, and hence only one mount
should be in the list.
* BuildKit doesn't own the mount directory, so should not try and remove
it on unmount.
Signed-off-by: Paul "TBBle" Hampson <Paul.Hampson@Pobox.com>
This is the shim used by the containerd Runtime V1 on Linux,
which was the default under containerd 1.2 and earlier.
Signed-off-by: Paul "TBBle" Hampson <Paul.Hampson@Pobox.com>
This reverts commit 9290c15ffc.
This was pinned during the upgrade to containerd 1.3 series, which
changed the default runtime on Linux to io.containerd.runc.v2.
No specific rationale was listed for this pinning, and clearly it's the
wrong thing to do in the presence of Windows, which does not have this
runtime.
Instead, we rely on the containerd-internal defaults, which distinguish
the runtimes for Linux and Windows.
Signed-off-by: Paul "TBBle" Hampson <Paul.Hampson@Pobox.com>
Tonis Tiigi
Alex Couture-Beil
Sebastiaan van Stijn
Paul "TBBle" Hampson
Akihiro Suda
Cory Bennett
Erik Sipsma