Commit Graph

275 Commits

Author SHA1 Message Date
Dr. Stefan Schimanski
b600e6c497 Rename KubernetesMesosScheduler.{Scheduler -> PodScheduler} 2015-11-12 01:33:50 +01:00
Dr. Stefan Schimanski
5db45baa8f Rename scheduler -> mesosScheduler in plugin_test 2015-11-12 01:33:50 +01:00
Dr. Stefan Schimanski
01a97ebc14 Rename KubernetesScheduler -> KubernetesMesosScheduler 2015-11-12 01:33:50 +01:00
Dr. Stefan Schimanski
12165cd6e0 Rename variable to ei for ExecutorInfo 2015-11-12 01:33:50 +01:00
Dr. Stefan Schimanski
4f13ba1af6 Revert "Merge pull request #15976 from mesosphere/sur-k8sm-475-error-checking"
The given merge will be rebased manually and appended to the scheduler refactoring.

This reverts commit 8d923afe23, reversing
changes made to d7458ddd4c.
2015-11-12 01:33:50 +01:00
Ravi L R
fe8658b041 Fix ports mapping in case of discontinuous port ranges in mesos offer
update elements in wildports directly instead of copy of the value
from wildports.
2015-11-11 12:28:26 -08:00
Sergiusz Urbaniak
9b7e40570e scheduler: fix shadowing in doSchedule 2015-11-10 20:54:58 +01:00
Sergiusz Urbaniak
6a16fe314f scheduler: remove err param from Register 2015-11-10 14:39:57 +01:00
Sergiusz Urbaniak
6fced095ce scheduler: simplify if-then-else handling in doSchedule 2015-11-10 14:39:57 +01:00
Sergiusz Urbaniak
c3311744c5 scheduler: correct doc in doSchedule 2015-11-10 14:39:56 +01:00
Sergiusz Urbaniak
71bd54adc6 scheduler: remove shadowed offer var in doSchedule
The variable `offer` is shadowed in the if block causing the `then`
branch to be non-effective.
2015-11-10 14:39:56 +01:00
Sergiusz Urbaniak
d53b716e97 scheduler: add newlines for readibility in doSchedule 2015-11-10 14:39:56 +01:00
Sergiusz Urbaniak
0ad0c1f2b1 scheduler: remove error param from doSchedule func
doSchedule currently accepts err values from previous invocation delegating
error handling in a location different from the caller which can be hard
to debug and is not a good practice.

We still maintain the same invariants after the refactoring. If an err
happened in a previous invocation to Register, the returned task object
was nil causing task.AcceptedOffer() to return false.

By not invoking doSchedule in case of an error we can eliminate the
first `err == nil` check in doScheduler.
2015-11-10 14:34:53 +01:00
Sergiusz Urbaniak
9bbfc35fa9 scheduler: add err check for procurement 2015-11-10 14:12:05 +01:00
Dr. Stefan Schimanski
9b00b7ef35 Avoid MockPodsListWatch deadlock of blocked channel while lock is hold 2015-11-05 13:20:29 +01:00
Daniel Smith
241551e36e set-gen example 2015-11-03 15:14:51 -08:00
k8s-merge-robot
9b953d75d2 Merge pull request #16111 from mesosphere/sttts-sandbox-overlay
Auto commit by PR queue bot
2015-11-03 03:00:57 -08:00
Dr. Stefan Schimanski
8fb77ce13e Exit scheduler if sandbox overlay does not exist 2015-11-03 11:27:41 +01:00
Dr. Stefan Schimanski
ed09e5c29d Add supported archive format for --mesos-sandbox-overlay 2015-11-03 11:27:41 +01:00
Dr. Stefan Schimanski
60ac1b932f Simplify serveFrameworkArtifact with library function for basenames 2015-11-03 11:27:41 +01:00
Dr. Stefan Schimanski
4062e2be2f Add executor sandbox overlay mechanism to distribute nsenter and socat 2015-11-03 11:27:40 +01:00
Wojciech Tyczynski
d47e21f19f Reuse TCP connections in Reflector between resync periods. 2015-10-26 19:35:25 +01:00
Steve Milner
8da2d4002f Fixes remaining imports per #15319 2015-10-20 07:13:13 -04:00
k8s-merge-robot
d329a66434 Merge pull request #15518 from ravilr/mesos_graceful_termination_tasklost
Auto commit by PR queue bot
2015-10-17 02:51:21 -07:00
James DeFelice
b28333ca43 enable profiling for the kubelet/executor when turned on at the scheduler 2015-10-15 21:22:41 +00:00
Ravi L R
99b03baa9b handle task lost status update from executor on docker daemon crash/restart by deleting the pod, so that the RC creates a new one, if the pod is backed by RC 2015-10-15 12:03:50 -07:00
Ravi L R
0e296c3dcd while deleting pod on terminal task status update, delete with grace_period=0 as the corresponding pod is already gone 2015-10-15 12:03:50 -07:00
eulerzgy
4ed5f9d240 adjust package name for pkg/controller directory 2015-10-12 09:51:27 +08:00
Dr. Stefan Schimanski
aea7985b85 Take default limits for cpu and mem into account in FitPredicate 2015-10-10 12:29:52 +02:00
Yu-Ju Hong
a3e60cc32e Rename imported package local name kubeletTypes to kubetypes
According to the naming guidelines, package name should not include mixedCaps.
2015-10-09 10:24:31 -07:00
Yu-Ju Hong
098ab05997 kubelet: move common types to kubelet/types
This would faciliate tasks such as moving code in pkg/kubelet to sub packages.
2015-10-08 14:38:01 -07:00
Alex Robinson
84f8f01d45 Merge pull request #14432 from mesosphere/sur-k8sm-492-launch-grace-period
MESOS make launchGracePeriod configurable
2015-10-05 13:27:12 -07:00
Sergiusz Urbaniak
42c84954fd executor,scheduler: make launchGracePeriod configurable 2015-10-02 07:43:49 +02:00
Dr. Stefan Schimanski
39942952cd Fix hostname check logic in scheduler 2015-10-01 06:50:43 +02:00
Brendan Burns
3474324d6a Merge pull request #14271 from mesosphere/sur-k8sm-441-flaky
MESOS: fix flaky TestPlugin_LifeCycle
2015-09-30 11:32:00 -07:00
Sergiusz Urbaniak
56fc0f5900 scheduler: reenable TestPlugin_LifeCycle, increase timeouts 2015-09-30 16:14:13 +02:00
eulerzgy
71b96422f4 set capital in some files 2015-09-30 14:46:20 +08:00
k8s-merge-robot
c807bea089 Merge pull request #13857 from mesosphere/node-labels
Auto commit by PR queue bot
2015-09-28 22:10:51 -07:00
Dr. Stefan Schimanski
67746908e5 Deleting gracefully terminating, not-scheduled pre-scheduled pods
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.
2015-09-26 23:42:08 +02:00
Dr. Stefan Schimanski
112f80fa4a Simplify FitPredicate for NodeSelector
We can assume that n != nil holds, because otherwise offers are skipped.
2015-09-26 10:36:49 +02:00
Dr. Stefan Schimanski
4d4ebe9f18 Add Mesos slave attributes as node labels
- 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
2015-09-26 09:46:56 +02:00
Dr. Stefan Schimanski
e4dcd97ac3 Dequeue pods in scheduler which are terminating 2015-09-22 16:41:43 +02:00
Dr. Stefan Schimanski
132f8892ed Move task-lost message prefix into messages module 2015-09-21 09:51:05 +02:00
Daniel Martí
586cb9126a Move pkg/util.Time to pkg/api/unversioned.Time
Along with our time.Duration wrapper, as suggested by @lavalamp.
2015-09-17 17:51:27 -07:00
Alex Mohr
bdfd469a85 Merge pull request #14137 from caesarxuchao/13823
rebase 13823
2015-09-17 15:25:52 -07:00
k8s-merge-robot
654dd38b1c Merge pull request #14120 from mesosphere/sttts-cleanup-hostname-override
Auto commit by PR queue bot
2015-09-17 14:23:36 -07:00
Kris
f4ad00d9ae Moving Status object to a common package 2015-09-17 14:09:53 -07:00
Dr. Stefan Schimanski
88466dd21e Cleanup --hostname-override logic in procurment 2015-09-17 18:41:16 +02:00
k8s-merge-robot
a204f31516 Merge pull request #13918 from mesosphere/reconcile-ignore-mirror-pods
Auto commit by PR queue bot
2015-09-17 02:36:31 -07:00
k8s-merge-robot
367e5571bf Merge pull request #13965 from mesosphere/healthz
Auto commit by PR queue bot
2015-09-17 02:13:31 -07:00