Design doc of 'Schedule DS Pod by default scheduler'. #1714
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
k8s-ci-robot
added
size/M
k8s-github-robot
added
the
sig/scheduling
|
That should be alpha in 1.10, as TaintNodeByCondition is still alpha right now. Any comments? |
7 tasks
|
/cc @kow3ns |
DStorck
reviewed
Feb 2, 2018
| @@ -0,0 +1,71 @@ | |||
| #Schedule DaemonSet Pods by default scheduler, not DaemonSet controller | |||
DStorck
reviewed
Feb 2, 2018
|
|
||
| * **Q**: Will this change/constrain update strategies, such as scheduling an updated pod to a node before the previous pod is gone? | ||
|
|
||
| **A**: nop, this will NOT change update strategies. |
DStorck
reviewed
Feb 2, 2018
|
|
||
| Before the discussion of solutions/options, there’s some requirements/questions on DaemonSet: | ||
|
|
||
| * **Q**: DaemonSet controller can make pods even the network of node is unavailable, e.g. CNI network providers (Calico, Flannel), Will this impact bootstrapping, such as in the case that a DaemonSet is being used to provide the pod network? |
DStorck
reviewed
Feb 2, 2018
|
|
||
| * **Q**: DaemonSet controller can make pods even the network of node is unavailable, e.g. CNI network providers (Calico, Flannel), Will this impact bootstrapping, such as in the case that a DaemonSet is being used to provide the pod network? | ||
|
|
||
| **A**: This will be handled in Support scheduling tolerating workloads on NotReady Nodes ([#45717](https://github.com/kubernetes/kubernetes/issues/45717)); after moving to check node’s taint, the DaemonSet pods will tolerant `NetworkUnavailable` taint. |
DStorck
reviewed
Feb 2, 2018
| Currently, pods of DaemonSet are created and scheduled by DaemonSet controller: | ||
|
|
||
| 1. DS controller filter nodes by nodeSelector and scheduler’s predicates | ||
| 2. For each nodes, create a Pod for it by setting spec.hostName directly; it’ll skip default scheduler |
DStorck
reviewed
Feb 2, 2018
| - dest_hostname | ||
| ``` | ||
|
|
||
| ##Reference |
DStorck
reviewed
Feb 2, 2018
| ##Reference | ||
|
|
||
| * [DaemonsetController can't feel it when node has more resources, e.g. other Pod exits](https://github.com/kubernetes/kubernetes/issues/46935) | ||
| * [DaemonsetController can't feel it when node recovered from outofdisk state](https://github.com/kubernetes/kubernetes/issues/46935) |
There was a problem hiding this comment.
wrong link, this is the same link as above , should be kubernetes/kubernetes#45628
k8s-github-robot
added
the
kind/design
bsalamat
reviewed
Feb 6, 2018
| * Hard to debug why DaemonSet’s Pod is not created, e.g. not enough resources; it’s better to have a pending Pods with predicates’ event | ||
| * Hard to support preemption in different components, e.g. DS and default scheduler | ||
|
|
||
| After [discussions](https://docs.google.com/document/d/1v7hsusMaeImQrOagktQb40ePbK6Jxp1hzgFB9OZa_ew/edit#), we come to a agreement that making DaemonSet to just produce pods like every other controller, and let them be scheduled by the regular scheduler, than to be its own scheduler. |
There was a problem hiding this comment.
s/a agreement/an agreement/
I would rephrase the sentence "we come to an agreement..." to:
SIG scheduling approved changing DaemonSet controller to create DaemonSet Pods and set their node-affinity and let them be scheduled by default scheduler. After this change, DaemonSet controller will no longer schedule DaemonSet Pods directly.
|
|
||
| * **Q**: DaemonSet controller can make pods even the network of node is unavailable, e.g. CNI network providers (Calico, Flannel), Will this impact bootstrapping, such as in the case that a DaemonSet is being used to provide the pod network? | ||
|
|
||
| **A**: This will be handled in Support scheduling tolerating workloads on NotReady Nodes ([#45717](https://github.com/kubernetes/kubernetes/issues/45717)); after moving to check node’s taint, the DaemonSet pods will tolerant `NetworkUnavailable` taint. |
|
|
||
| * **Q**: DaemonSet controller can make pods even when the scheduler has not been started, which can help cluster bootstrap. | ||
|
|
||
| **A**: As the scheduling logic is moved to default scheduler, the kube-scheduler is required after this proposal. |
There was a problem hiding this comment.
s/the kube-scheduler is required after this proposal/the kube-scheduler must be started during cluster start-up/
|
|
||
| This option is to leverage NodeAffinity feature to avoid introducing scheduler’s predicates in DS controller: | ||
|
|
||
| 1. DS controller filter nodes by nodeSelector, but did NOT check against scheduler’s predicates (e.g. PodFitHostResources) |
| This option is to leverage NodeAffinity feature to avoid introducing scheduler’s predicates in DS controller: | ||
|
|
||
| 1. DS controller filter nodes by nodeSelector, but did NOT check against scheduler’s predicates (e.g. PodFitHostResources) | ||
| 2. For each nodes, DS controller creates a Pod for it with following NodeAffinity |
There was a problem hiding this comment.
s/each nodes/each node/
s/with following/with the following/
| 1. DS controller filter nodes by nodeSelector, but did NOT check against scheduler’s predicates (e.g. PodFitHostResources) | ||
| 2. For each nodes, DS controller creates a Pod for it with following NodeAffinity | ||
| 3. When sync Pods, DS controller will map nodes and pods by this NodeAffinity to check whether Pods are started for nodes | ||
| 4. In scheduler, the Daemon pods will keep pending if predicates failed, e.g. PodFitHostResources; for critical daemons, DS controller will create Pods with critical pods annotation and leverage scheduler/kubelet’s logic to handle it; similar practice to [priority/preemption](https://github.com/kubernetes/features/issues/268 ) |
There was a problem hiding this comment.
As a part of enabling priority and preemption, we must ensure that all critical DaemonSet Pods get an appropriate critical priority. If they have critical priority, scheduler will ensure that they will be scheduled even when the cluster is under resource pressure. Scheduler preempts other Pods in such condition to schedule critical Pods.
k82cn
force-pushed
the
k8s_548
branch
2 times, most recently
from
6597e10 to
f936c42
Compare
bsalamat
reviewed
Feb 8, 2018
| 1. DS controller filter nodes by nodeSelector, but does NOT check against scheduler’s predicates (e.g. PodFitHostResources) | ||
| 2. For each node, DS controller creates a Pod for it with the following NodeAffinity | ||
| 3. When sync Pods, DS controller will map nodes and pods by this NodeAffinity to check whether Pods are started for nodes | ||
| 4. In scheduler, the Daemon pods will keep pending if predicates failed, e.g. PodFitHostResources; for critical daemons, |
There was a problem hiding this comment.
s/the Daemon pods will keep pending if predicates failed, e.g. PodFitHostResources; for critical daemons,/Daemon Pods will stay pending if scheduling predicates fail. To avoid this,/
Signed-off-by: Da K. Ma <[email protected]>
|
/lgtm |
k8s-ci-robot
added
the
lgtm
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: bsalamat, k82cn The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these OWNERS Files:
Approvers can indicate their approval by writing |
k8s-ci-robot
added
the
approved
8 tasks
MadhavJivrajani
pushed a commit
to MadhavJivrajani/community
that referenced
this pull request
Nov 30, 2021
Design doc of 'Schedule DS Pod by default scheduler'.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
xref kubernetes/enhancements#548
/cc @bgrant0607 , @bsalamat , @kubernetes/sig-apps-feature-requests