Commit Graph

3365 Commits

Author SHA1 Message Date
Wei Huang
ff831c3df2 EvenPodsSpread: Supports Preemption (addPod)
- add addPod() for podSpreadMap
2019-07-24 15:27:55 -07:00
Wei Huang
9e9808d0ab EvenPodsSpread: Supports Preemption (removePod)
- add removePod() for podSpreadMap
2019-07-24 15:27:54 -07:00
Kubernetes Prow Robot
acd597dbac Merge pull request #80305 from hex108/cleanup_score_plugin
Precheck score plugins' weight when initializing
2019-07-24 15:20:05 -07:00
Kubernetes Prow Robot
d83cf5f274 Merge pull request #77828 from Huang-Wei/eps-pred-core
Even Pods Spread - 3. Predicates Core
2019-07-24 15:19:52 -07:00
Aldo Culquicondor
8e9af0185d Remove NodeLister from Scheduler Configurator
Signed-off-by: Aldo Culquicondor <acondor@google.com>
2019-07-24 17:16:30 -04:00
Kubernetes Prow Robot
cbae6ed036 Merge pull request #80220 from alculquicondor/refactor/scheduler-cache
Use scheduler cache in affinity priority functions
2019-07-24 13:50:06 -07:00
Wei Huang
1822085088 EvenPodsSpread: update 'selfMatch' logic 2019-07-24 10:28:46 -07:00
Wei Huang
39e459ae9a fixup: address comments 2019-07-24 10:28:46 -07:00
Wei Huang
b99fb9187b EvenPodsSpread: UT on genericScheduler.Schedule() 2019-07-24 10:28:46 -07:00
Wei Huang
e0e3889d74 EvenPodsSpread: Core Predicate logic 2019-07-24 10:28:46 -07:00
Wei Huang
08e7b3bacb EvenPodsSpread: Define a new Predicate 2019-07-24 10:28:46 -07:00
draveness
7549c53a38 fix: return empty string when status is nil 2019-07-24 20:55:11 +08:00
Aldo Culquicondor
f58abdf966 Use scheduler cache in affinity priority functions
Make the cache implement NodeLister and expose it to the priority
functions. This way, the priority functions make use of a single cache,
the scheduler's, instead of mixing it with the lister's caches.

Signed-off-by: Aldo Culquicondor <acondor@google.com>
2019-07-23 22:10:49 -04:00
Shintaro Murakami
c94e800008 Use v1helper.GetPersistentVolumeClaimClass for compatibility 2019-07-24 10:35:00 +09:00
Jun Gong
a3ed664246 Add doc that plugins in binding cycle should not use NodeInfoSnapshot() 2019-07-23 20:25:39 +08:00
Jun Gong
df14adf474 Send a reject message to permit plugin when preempting a WaitingPod 2019-07-23 10:23:54 +08:00
Wei Huang
f822487f05 EvenPodsSpread: match selector of each constraint independently
- see more discussion at https://github.com/kubernetes/kubernetes/pull/77760#discussion_r305107973
2019-07-22 17:51:23 -07:00
Wei Huang
c23aef40f2 [eps-pred-meta] auto-gen files 2019-07-22 17:34:56 -07:00
Wei Huang
249752cc1f fixup: fix comments and use a channel to pass err 2019-07-22 17:34:56 -07:00
Wei Huang
dce6686c9a EvenPodsSpread: refactor "chained" utils
- move "chanined" utils to pkg/scheduler/testing/utils.go so as to be re-used by all scheduler tests
2019-07-22 17:34:56 -07:00
Wei Huang
3dbef991a3 EvenPodsSpread: refactor topologyPairsPodSpreadMap
- update minMatchMap from []int32 to map[string]int32
2019-07-22 17:34:56 -07:00
Wei Huang
a242e376ac EvenPodsSpread: PredicateMetadata initilization
- build a new `topologyPairsPodSpreadMap` into PredicateMetadata
- update ShallowCopy()
- unit tests
2019-07-22 17:34:56 -07:00
Kubernetes Prow Robot
f31d786927 Merge pull request #78477 from YoubingLi/filter
Fixes 78001 The implementation of Filter extension for the new framework
2019-07-22 16:48:53 -07:00
Shintaro Murakami
169537499c Record metrics when framework returns unschedulable error 2019-07-22 16:01:22 +09:00
Chun Chen
1f3bc52eab Calling Unreserve plugin before recordSchedulingFailure in case of binding failure
Calling recordSchedulingFailure puts the pod back to scheduling queue in another
goroutine so pod may get a chance to be bond again before unreseve plugin cleaning
state about it.
2019-07-19 10:55:06 +08:00
Jun Gong
2dc5cf8c43 Set score plugin's weight to 1 if it is not set 2019-07-18 22:35:34 +08:00
Jun Gong
66347b516c Precheck score plugins' weight when initializing 2019-07-18 21:19:21 +08:00
Kubernetes Prow Robot
ce80aeaac5 Merge pull request #78319 from egernst/scheduler-changes
Scheduler changes to introduce alpha support for Pod Overhead
2019-07-18 01:52:11 -07:00
Youbing Li
7f9dd949e0 The implementation of Filter extension for the new framework 2019-07-17 03:28:30 -04:00
Eric Ernst
9babbf8bd7 pod-overhead: autogenerated code for scheduling changes
Signed-off-by: Eric Ernst <eric.ernst@intel.com>
2019-07-16 12:56:19 -07:00
Eric Ernst
62e9717386 scheduler: add pod Overhead support for requests
Signed-off-by: Eric Ernst <eric.ernst@intel.com>
2019-07-16 12:56:19 -07:00
Abdullah Gharaibeh
a80425bd23 Fixed build files. 2019-07-16 10:24:10 -04:00
Abdullah Gharaibeh
73704f1b8e Use error channel to capture first error. 2019-07-16 10:00:42 -04:00
Abdullah Gharaibeh
c54c4d1962 Score plugin for the scheduling framework. 2019-07-16 09:38:37 -04:00
Kubernetes Prow Robot
a155f3b0e9 Merge pull request #80132 from draveness/feature/rename-register-priority-function
fix: rename RegisterPriorityFunction2 to RegisterPriorityMapReduceFunction
2019-07-15 20:27:48 -07:00
Kubernetes Prow Robot
5a32bea904 Merge pull request #79777 from draveness/feature/refactor-scheduling-predicates
feat: use channel instead of mutex in scheduling predicates
2019-07-15 20:27:35 -07:00
Chun Chen
e04f03d3cb Fix flaky test TestBindPlugin 2019-07-15 23:36:06 +08:00
Chun Chen
3bb1a081ab Revert "Revert "Add Bind extension point of the scheduling framework""
This reverts commit b8950dab50.
2019-07-15 23:36:02 +08:00
draveness
a4618d6790 fix: rename RegisterPriorityFunction2 to RegisterPriorityMapReduceFunction 2019-07-15 08:17:52 +08:00
Kubernetes Prow Robot
7285829b83 Merge pull request #78263 from misterikkit/nonZeroDocs
Add documentation about "non-zero requests"
2019-07-14 13:19:07 -07:00
draveness
fab1948a1b feat: use channel instead of mutex in scheduling predicates 2019-07-13 09:11:52 +08:00
Yassine TIJANI
08522f8e5a move scheduler to use v1beta1.events
Signed-off-by: Yassine TIJANI <ytijani@vmware.com>
2019-07-12 11:39:01 +02:00
jingxueli
6a1bbff60c keep processing other nodes for the nil node error 2019-07-11 15:34:40 +08:00
Kubernetes Prow Robot
b7757ffbb9 Merge pull request #78280 from sevensees/master
Trace step should be at the end of calls rather than the begin.
2019-07-10 21:39:07 -07:00
Kubernetes Prow Robot
79564ebff7 Merge pull request #79877 from draveness/feature/use-scheduler-cache-instead
fix: use schedulerCache instead of podlister in config factory
2019-07-10 16:24:40 -07:00
Kubernetes Prow Robot
9e5018ba20 Merge pull request #79453 from chendave/short-circuit
Short-circuit the evaluation of `cpuFraction` and `memoryFraction`
2019-07-10 16:24:27 -07:00
shiyan2016
c329d5117d cancel process node if error occurs 2019-07-09 19:50:48 -07:00
Kubernetes Prow Robot
6653a490fb Merge pull request #79532 from snowplayfire/update-errMsg
improve error msg for predicate meta data
2019-07-09 13:46:37 -07:00
Jianfei Bai
95bd52cfe9 Trace step should be at the end of calls rather than the begin. 2019-07-08 16:54:04 +08:00
draveness
2fe0d4da27 fix: use schedulerCache instead of podlister in config factory 2019-07-08 14:35:30 +08:00