This tag of hcsshim brings in a couple welcome features/improvements. One being
exposing a way to query for hns endpoint statistics (Packets received/sent etc.).
This tag also contains some optimizations for querying whether a certain HCN feature
is supported, which is a common workflow in kube-proxy on Windows. The first result
from querying HCN is now cached so further calls can skip the hcn query as well as the
version range parsing that was performed. This also gets rid of some redundant logs
that used to hit everytime the version range parsing occurred.
The Go-winio dep bump, and all of the ctrd deps are transitive only. Nothing new is needed/intended
to be used.
Signed-off-by: Daniel Canter <dcanter@microsoft.com>
This replaces the experimental logr v0.4 with the stable v1.1.0
release. This is a breaking API change for some users because:
- Comparing logr.Logger against nil is not possible anymore:
it's now a struct instead of an interface. Code which
allows a nil logger should switch to *logr.Logger as type.
- Logger implementations must be updated in lockstep.
Instead of updating the forked zapr code in json.go, directly using
the original go-logr/zapr is simpler and avoids duplication of effort.
The updated zapr supports logging of numeric verbosity. Error messages
don't have a verbosity (= always get logged), so "v" is not getting
added to them anymore.
Source code logging for panic messages got fixed so that it references
the code with the invalid log call, not the json.go implementation.
Finally, zapr includes additional information in its panic
messages ("zap field", "ignored key", "invalid key").
A pod that has been rejected by admission will have status manager
set the phase to Failed locally, which make take some time to
propagate to the apiserver. The rejected pod will be included in
admission until the apiserver propagates the change back, which
was an unintended regression when checking pod worker state as
authoritative.
A pod that is terminal in the API may still be consuming resources
on the system, so it should still be included in admission.
So that the width estimate has some effect but not a grossly excessive
one.
Added the fifo::Peek method to simplify the fifo client code.
Also renamed the queueSet::estimatedServiceTime field to
estimatedServiceSeconds to make the units clear.
In older versions of Kubernetes (at least pre-0.19, it's the earliest
this test will run unmodified on), Pods that depended on devices could be
restarted after the device plugin had been removed. Currently however,
this isn't possible, as during ContainerManager.GetResources(), we
attempt to DeviceManager.GetDeviceRunContainerOptions() which fails as
there's no cached endpoint information for the plugin type.
This commit therefore breaks apart the existing test into two:
- One active test that validates that assignments are maintained across
restarts
- One skipped test that validates the behaviour after GPUs have been
removed, in case we decide that this is a bug that should be fixed in
the future.