Automatic merge from submit-queue (batch tested with PRs 39538, 40188, 40357, 38214, 40195)
test for host cleanup in unfavorable pod deletes
addresses issue #31272 with a new e2e test in _kubelet.go_
```release-note
NONE
```
The test checks the individual kubelet /runningPods endpoint based on the
initial list of nodes it observes. It is important that all pods are
scheduled only onto those nodes. Apply node labels to ensure no stray pods on
other nodes.
This scopes down the initially ambitious PR:
https://github.com/kubernetes/kubernetes/pull/14960 to replace just
`pause` and `fluentd-elasticsearch` to come through `beta.gcr.io`.
The v2 versions have been pushed under new tags, `pause:2.0` and
`fluentd-elastisearch:1.12`.
NOTE: `beta.gcr.io` will still serve images using v1 until they are repushed with v2. Pulls through `gcr.io` will still work after pushing through `beta.gcr.io`, but will be served over v1 (via compat logic).
This test tracks kubelet resource usage over a long period of time (1hr)
when running N pods (e.g., N=0,50), and prints out the resource usage. This
would give us an idea how much kubelet's management overhead is in a stable
cluster.
Some followup items:
* Use a more realistic workload (e.g., including probing)
* Fail the test if the resource usage is too high.
Caveat:
* We assume the scheduler would do a decent job distributing the pause pods,
but we should double check.
* Cluster addon pods could be unevenly distributed and skews the resource
usage on nodes.
Other cluster provider than gce or gke might have different cgroup layouts.
From outside we cannot know how these look like (especially in conformance test
which do not know the cluster provider at all).
Hence, this PR defaults to only the "/" cgroup to collect stats for. In the case
of gce or gke the full container list is tested.
Fixes https://github.com/mesosphere/kubernetes-mesos/issues/436
A lot of packages use StringSet, but they don't use anything else from
the util package. Moving StringSet into another package will shrink
their dependency trees significantly.
This change adds resourceMontior, which spawns a goroutine per node to poll the
container stats for known, relevant containers, computes the resource usage and
stores the data. Users can then examine the data in the buffer to get resource
of each individual containers.
This change adds simple utility functions to query kubelet for the resource
usage of major containers (e.g. /kubelet, /docker-daemon, etc) for the past
time interval (up to 2 minutes). This is intended for spot checking of the
recent resource usage on the node.
The test verifies that kubelet deletes the pods/containers within a reasonable
time. It queries the kubelet /runningpods endpoint to retrieve a list of
running pods directly. The outline of the test is:
- Create an RC
- Wait until all pods are running (based on the pod status)
- Verify pods are running by querying the /runningpods
- Delete the RC
- Check all pods are deleted by querying /runningpods
Michelle Au