manager change in commit 7b1ae66.
Also changes the test to make sure node is indeed ready after Kubelet
restart. The previous readiness check may use old API state but
didn't run into the issue due to the delay of waiting for pod restart.
Kubelet restart process seems to get a bit slower recently. From running
the gpu_device_plugin e2e_node test on GCE, I saw it took ~37 seconds
for kubelet to start CM DeviceManager after it restarts, and then took
~12 seconds for the gpu device plugin to re-register. As the result,
this e2e_node test fails because the current 10 sec waiting time is too
small. Restarting a container also seems to get slower that it sometimes
exceeds the current 2 min waiting time in ensurePodContainerRestart().
This change increase both waiting time to 5 min to leave enough space
on slower machines.
incompatible changes:
- Add GetDevicePluginOptions rpc call. This is needed when we switch
from Registration service to probe-based plugin watcher.
- Change AllocateRequest and AllocateResponse to allow device requests
from multiple containers in a pod. Currently only made mechanical
change on the devicemanager and test code to cope with the API but
still issues an Allocate call per container. We can modify the
devicemanager in 1.11 to issue a single Allocate call per pod.
The change will also facilitate incremental API change to communicate
pod level information through Allocate rpc if there is such future
need.
David Ashpole