In upstream the kubelet is responsible for all pods which have the spec.NodeName
set. In Mesos we have a two-stage scheduling process:
1. pods with a pre-set spec.NodeName are still scheduled by the scheduler.
2. The kubelet will only see them when a Mesos task was started and the executor
passes the pod to the kubelet.
With this PR a pod with spec.NodeName which is gracefully terminated, but not
yet scheduled, e.g.
- because the termination happened just after creation and the scheduler was
not fast enough
- because the NodeSelector does not match
is deleted by the Mesos scheduler.
- pre-create node api objects from the scheduler when offers arrive
- decline offers until nodes a registered
- turn slave attributes as k8s.mesosphere.io/attribute-* labels
- update labels from executor Register/Reregister
- watch nodes in scheduler to make non-Mesos labels available for NodeSelector matching
- add unit tests for label predicate
- add e2e test to check that slave attributes really end up as node labels
- remove bleeding of registry-internal objects, without any locking
- rename from SlaveStorage to Registry which fits much better to what
it actually does
- new: introduce AllocationStrategy, Predicate, and Procurement to scheduler pkg
- new: --contain-pod-resources flag (workaround for docker+systemd+mesos problems)
- new: --account-for-pod-resources flag (for testing overcommitment)
- bugfix: forward -v flag from minion controller to executor
- move assigned slave to T.Spec.AssignedSlave
- only create the BindingHost annoation in prepareTaskForLaunch
- recover the assigned slave from annotation and write it back to the T.Spec field
Before this patch the annotation were used to store the assign slave. But due
to the cloning of tasks in the registry, this value was never persisted in the
registry.
This patch adds it to the Spec of a task and only creates the annotation
last-minute before launching.
Without this patch pods which fail before binding will stay in the registry,
but they are never rescheduled again. The reason: the BindingHost annotation does
not exist in the registry and not on the apiserver (compare reconcilePod function).
Before NodeName in the pod spec was used. Hence, pods with a fixed, pre-set
NodeName were never scheduled by the k8sm-scheduler, leading e.g. to a failing
e2e intra-pod test.
Fixesmesosphere/kubernetes-mesos#388
This patch
- set limits (0.25 cpu, 64 MB) on containers which are not limited in pod spec
(these are also passed to the kubelet such that it uses them for the docker
run limits)
- sums up the container resource limits for cpu and memory inside a pod,
- compares the sums to the offered resources
- puts the sums into the Mesos TaskInfo such that Mesos does the accounting
for the pod.
- parses the static pod spec and adds up the resources
- sets the executor resources to 0.25 cpu, 64 MB plus the static pod resources
- sets the cgroups in the kubelet for system containers, resource containers
and docker to the one of the executor that Mesos assigned
- adds scheduler parameters --default-container-cpu-limit and
--default-container-mem-limit.
The containers themselves are resource limited the Docker resource limit which
the kubelet applies when launching them.
Fixesmesosphere/kubernetes-mesos#68 and mesosphere/kubernetes-mesos#304
Sergiusz Urbaniak