KEP-1397: descheduler integration with evacuation API as an alternative to eviction API #1354

ingvagabund · 2024-02-26T09:59:14Z

Is your feature request related to a problem? Please describe.

The descheduler eviction policy is built on top of the eviction API. The API currently does not support eviction requests that are not completed right away. Instead, any eviction needs to either succeed or be rejected in response. Nevertheless, there are cases where an eviction request is expected to only initiate eviction. While getting confirmation or rejection of the eviction initiation (or its promise).

Describe the solution you'd like

Utilize evacuation API as an alternative to eviction API. As an interim solution (until the evacuation API is available) allow to interpret pods with a special annotation as a request for eviction initiation instead of expecting an immediate eviction.

Resolves: #1397

songtao98 · 2024-05-11T09:04:49Z

Hi, @ingvagabund . I have some questions about how does this KEP works in Descheduling Framework.

Do you expect to use this new eviction handling mechanism to replace existing PodEvictor （who call Eviciton API directly）, so that all users evict their pods through it?
Do you consider to take one more step to make it as a extension point, e.g. Evict, so that users can decide which way to evict their pods, and existing PodEvictor or your extension by this KEP can both be implemented as a EvictPlugin(different with existing EvictorPlugin)?

ingvagabund · 2024-05-11T18:49:08Z

@songtao98 thank you for asking the right questions.

Do you expect to use this new eviction handling mechanism to replace existing PodEvictor （who call Eviciton API directly）, so that all users evict their pods through it?

Eventually, that's the ultimate goal. However, the first implementations will provide both eviction API and evacuation API. So we have some soft time for early adopters to share their experience and corner case for which evacuation API does not work/scale well. Followed by discussions with evacuation API designers about possible resolutions before the new API goes GA.

Do you consider to take one more step to make it as a extension point, e.g. Evict, so that users can decide which way to evict their pods, and existing PodEvictor or your extension by this KEP can both be implemented as a EvictPlugin(different with existing EvictorPlugin)?

Not right now. Ultimately, the new evacuation API should be a special case of the eviction API. The details are yet to be discussed.

songtao98 · 2024-05-13T01:59:02Z

thank you for asking the right questions.

@ingvagabund Thank you for your reply! For question 1, the strategy sounds great. For question 2, I'm not sure if someone has ever proposed a custom Evictor mechanism(to customize the real evictor like PodEvictor, not currently EvictorPlugin), but we think this capability will eventually be meaningful and already implemented one. We are also looking forward to more discussion about details, and if the community think it's the right way, we will be more than happy to contribute : )

ingvagabund · 2024-05-13T08:34:52Z

We think this capability will eventually be meaningful and already implemented one. We are also looking forward to more discussion about details, and if the community think it's the right way, we will be more than happy to contribute : )

Few sig calls back we discussed the possibility of writing mini-keps. Turning the low level eviction mechanism into a plugin is a good example for it :) If there are good use cases where a user can benefit from a custom eviction mechanism. Would you be willing to write a proposal for this? The current proposal focuses mainly on integration with the new evacuation API (and annotation as an interim solution). Pluginazing the evictor itself is related, yet it deserves its own attention.

songtao98 · 2024-05-14T02:13:33Z

Few sig calls back we discussed the possibility of writing mini-keps. Turning the low level eviction mechanism into a plugin is a good example for it :) If there are good use cases where a user can benefit from a custom eviction mechanism. Would you be willing to write a proposal for this? The current proposal focuses mainly on integration with the new evacuation API (and annotation as an interim solution). Pluginazing the evictor itself is related, yet it deserves its own attention.

For sure! We are certainly willing to draft a proposal for customizing the eviction mechanism. However, it might take some time as I need to organize some details and perform the necessary abstractions. BTW, regarding the mini-keps you mentioned, are there any specific format requirements? If not, I will follow the current KEP template.

songtao98 · 2024-05-14T03:21:07Z

keps/1397-evictions-in-background/README.md

+|**Pod nomination and sorting**|Preferring pods with both `descheduler.alpha.kubernetes.io/request-evict-only` and `descheduler.alpha.kubernetes.io/eviction-in-progress` annotations.|Preferring pods that have a corresponding evacuation CR present.|
+|**Cache reconstruction**|When a descheduler gets restarted and a new internal caches (with tracked annotations) cleared the descheduler lists all pods and populate the cache with any pod that has both annotation (`descheduler.alpha.kubernetes.io/request-evict-only` and `descheduler.alpha.kubernetes.io/eviction-in-progress`) present. In case only the first annotation is present but the eviction request was already created, the update event handler will catch the second annotation addition and the cache gets synced. In the worst case the limit of max number of pods evicted gets exceeded.|The descheduler waits until all evacution CRs are synced.|
+
+Given the evucation API is still a work in progress without any existing implementation, the annotation based eviction can be seen as an v1alpha1 implemenation of this proposal. Integration with the evacuation API as an v1alphaN or beta implementation.


Given the evucation API is still a work in progress without any existing implementation,

Typo here "evucation"

…ve to eviction API

knelasevero · 2024-06-05T17:28:05Z

In Filip's proposal (I need to read it in more detail, but I skimmed it already), there is this notion for Evacuator Prioritization and multiple evacuators (and a lot of other details not considered here). To me it sounds like when going from alpha implementation (here) to beta, and trying to assimilate the other complexities from the real evacuator API too much would change to call the graduation here a graduation from alpha to beta? Maybe it would be another alpha for a while before becoming beta? (or, again, as above comment, it would be worth to consider all those things from the start...)

knelasevero · 2024-06-05T17:33:27Z

keps/1397-evictions-in-background/README.md

+- Evacuation API based evictions
+- Identifying corner cases that are not addressed by the evacuation API
+- `descheduler.alpha.kubernetes.io/request-evict-only` annotation is deprecated
+- Workloads do not set `descheduler.alpha.kubernetes.io/request-evict-only` annotation


@atiratree then also implementation around prioritization, detailed status reporting, and cancellation policies as per the Evacuation API? Then, not a beta graduation, right?

I suppose, second alpha would be more appropriate.

Also, Workloads do not set descheduler.alpha.kubernetes.io/request-evict-only annotation is not actionable. Maybe s/do/should/ and document this?

The implementation will not get promoted to beta without the evacuation API integration. We will be rolling at least two v1alpha versions.

Workloads do not set descheduler.alpha.kubernetes.io/request-evict-only annotation is not actionable. Maybe s/do/should/ and document this?

What about Workloads are not expected to set descheduler.alpha.kubernetes.io/request-evict-only annotation when deprecated?

knelasevero · 2024-06-05T17:34:04Z

keps/1397-evictions-in-background/README.md

+      2. Error code, resp. response text is checked to distinguish between
+         a genuine error and a confirmation of an eviction in background.
+
+##### Evacuation API (v1alphaN or beta)


Oh, i see v1alphaN is considered, I missed this

Maybe more than considered it should be a must given the expected changes

Changed to "v1alphaN and beta"

knelasevero · 2024-06-05T17:36:56Z

This looks great and is very detailed. I need to read both proposals a bit more deeply to give more insights, but wanted to give my initial thoughts so the KEP can go forward

atiratree · 2024-06-12T20:21:38Z

keps/1397-evictions-in-background/README.md

+      annotation or confronting the internal caches.
+   2. Each descheduling plugin nominates a set of pods to be evicted.
+   3. All the nominated pods are sorted based on the eviction-in-progress first priority.
+4. Pod eviction: the descheduler starts evicting nominated pods until a limit is reached


Is the eviction going to be handled in each plugin separately as it is currently?

What are the guarantees that a plugin will use the Evacuator/Evictor correctly? E.g. calling pre-checks?

Is the eviction going to be handled in each plugin separately as it is currently?

Yes. The current mechanism will be persisted.

What are the guarantees that a plugin will use the Evacuator/Evictor correctly? E.g. calling pre-checks?

None. Each plugin is given a handler for evicting/evacuating a pod. No pod will be aware whether an eviction or evacuation API is invoked under the hood.

By pre-checks you mean excluding pods with already existing evacuation requests?

atiratree · 2024-06-12T20:32:48Z

keps/1397-evictions-in-background/README.md

+##### Evacuation API (v1alphaN or beta)
+
+1. Policy configuration
+   1. The cluster administrator configures the descheduler to enable the new functionality.


What is the descheduler kubernetes version backwards compatibility? There should be probably a check if the Evacuation API resource exists in the cluster

We run a check for whether an eviction API is supported in

descheduler/pkg/descheduler/descheduler.go

Line 250 in c7f8670

evictionPolicyGroupVersion, err := eutils.SupportEviction(rs.Client)

. I expect a similar check for the evacuation API.

atiratree · 2024-06-12T20:34:23Z

keps/1397-evictions-in-background/README.md

+      all the internal counters to take into account all pods that are subjects
+      to background eviction. By listing evacuation requests or confronting the internal caches.
+   2. Each descheduling plugin nominates a set of pods to be evicted.
+   3. All the nominated pods are sorted based on the eviction-in-progress first priority.


note: eviction in progress is recognized by a presence of Evacuation object

Ack. The descheduler is expected to read all existing evacuation objects/CRs and classify corresponding pods as "evacuation requested or in progress".

atiratree · 2024-06-12T20:35:52Z

keps/1397-evictions-in-background/README.md

+      to background eviction. By listing evacuation requests or confronting the internal caches.
+   2. Each descheduling plugin nominates a set of pods to be evicted.
+   3. All the nominated pods are sorted based on the eviction-in-progress first priority.
+4. Pod eviction: the descheduler creates an evacuation request for each nominated pod until a limit is reached


implementation detail, but

it should add a evacuation.coordination.k8s.io/instigator_${EVACUATION_INSTIGATOR_SUBDOMAIN} finalizer to the Evacuation when creating it

if the Evacuation exists it should ensure the finalizer is present

atiratree · 2024-06-12T20:53:07Z

keps/1397-evictions-in-background/README.md

+- The upstream eviction API does not currently support evictions in background
+  implemented by external components. For that, there's no community
+  dedicated error code nor a response text for this type of eviction.
+- The eviction in background error code will be temporary mapped to 429 (TooManyRequests).


Ok, I see. No retries.

There is also a standing issue about the confusion of today's Eviction API request errors: kubernetes/kubernetes#106286. I am not sure if it is a good idea to add another behavior dependent on external component to that. At least having a distinguished code could make it more manegable

atiratree · 2024-06-12T20:59:31Z

keps/1397-evictions-in-background/README.md

+- Evacuation API based evictions
+- Identifying corner cases that are not addressed by the evacuation API
+- `descheduler.alpha.kubernetes.io/request-evict-only` annotation is deprecated
+- Workloads do not set `descheduler.alpha.kubernetes.io/request-evict-only` annotation


I suppose, second alpha would be more appropriate.

Also, Workloads do not set descheduler.alpha.kubernetes.io/request-evict-only annotation is not actionable. Maybe s/do/should/ and document this?

atiratree · 2024-06-12T21:00:47Z

keps/1397-evictions-in-background/README.md

+- Feature implemented behind a feature flag
+- Initial e2e tests completed and enabled
+- Annotation based evictions
+- Workloads do not set `descheduler.alpha.kubernetes.io/request-evict-only` annotation


a bit confusing to see the deprecation in alpha

That depends on how many alphas we will have. The annotation gets deprecated the moment the evacuation API is implemented as alpha. While we will need to wait until the evacuation API gets promoted to beta before we can promote the eviction-in-background to beta.

atiratree · 2024-06-12T21:35:20Z

keps/1397-evictions-in-background/README.md

+
+```go
+type PodEvacuator interface {
+    EvacutePod(ctx context.Context, pod *v1.Pod) error


what guarantee we have each plugin will call this interface? Isn't it better to override/change the Evictor interface?

The Evictor interface will get replaced by PodEvacuator interface under the hood. Currently, all plugins are expected to invoke <plugin>.handle.Evictor().Evict method. The method will invoke the evacuation API once available and get renamed to a more suitable name.

atiratree · 2024-06-12T21:38:17Z

keps/1397-evictions-in-background/README.md

+* The evacuation API proposal mentions more than a one entity can request an eviction.
+  Should the descheduler take these into account as well?
+  What if another entity decides to evict a pod that is of a low priority
+  from the descheduler's point of view?
+  **Answer**: The first implementation will consider all pods with an existing
+  evacuation request as "already getting evacuated". Independent of the pod priorities.


The important part to decide for the descheduler is, if it wants to subscribe to evacuating these lower evacuations as well. This is done by setting the finalizer and prevents the evacuation cancellation.

Good point. This depends on how a user will see these evacuations. We could introduce a new option for this case so the descheduler sets the finalizer even for these pods and count them as "marked for eviction by the descheduler".

Have you considered a concept of a maximal number of pods that can be evacuated per time? A global setting for an entire cluster?

atiratree · 2024-06-12T21:47:22Z

keps/1397-evictions-in-background/README.md

+     with `descheduler.alpha.kubernetes.io/eviction-in-progress` or creating evacuation requests).
+  2. Run the descheduler and observe only an expected amount of pods
+     is requested to be evicted/evacuated.
+- `EvacuationTTL` for testing an evacuation request (with a single owner


I am a bit confused, do we introduce the TTL for the first version or not?

I see I have mentioned the TTL will not be implemented in the first version under Open Questions. Yet, v1alpha1 section mentions it will. I removed that part from v1alpha1.

v1alpha1 will not introduce the TTL mechanism. Yet, it might get implemented in later v1alpha versions.

atiratree · 2024-06-12T21:49:56Z

Do we consider the alpha implementation here to be an workaround until the until the evacuation API is fully integrated (in the future, in beta)?

I understand trying to deliver value earlier and then already start to get feedback, but have you checked with Feature Requesters and stakeholders if that is really what they want? Maybe it makes sense to start with the evacuation API integration from the start.

FIY, we are still discussing the Evacuation API feature and it was deferred from 1.31

ingvagabund · 2024-06-17T13:26:28Z

Do we consider the alpha implementation here to be an workaround until the until the evacuation API is fully integrated (in the future, in beta)?

Yes

I understand trying to deliver value earlier and then already start to get feedback, but have you checked with Feature Requesters and stakeholders if that is really what they want? Maybe it makes sense to start with the evacuation API integration from the start.

I did. This is an important feature for KubeVirt. There's a strong interest to have this feature merged sooner than later. Waiting until the Evacuation API is implemented/alpha would be the best course of actions. Yet, it's not a blocker for implementing the first alpha of this feature.

ingvagabund · 2024-06-17T13:28:25Z

In Filip's proposal (I need to read it in more detail, but I skimmed it already), there is this notion for Evacuator Prioritization and multiple evacuators (and a lot of other details not considered here). To me it sounds like when going from alpha implementation (here) to beta, and trying to assimilate the other complexities from the real evacuator API too much would change to call the graduation here a graduation from alpha to beta? Maybe it would be another alpha for a while before becoming beta? (or, again, as above comment, it would be worth to consider all those things from the start...)

Currently, the promotion to beta requires implementation of the evacuation API. Thus, we need to stay aligned with it and can not promote to beta until the evacuation API is beta.

knelasevero · 2024-06-17T17:26:29Z

Do we consider the alpha implementation here to be an workaround until the until the evacuation API is fully integrated (in the future, in beta)?

Yes

I understand trying to deliver value earlier and then already start to get feedback, but have you checked with Feature Requesters and stakeholders if that is really what they want? Maybe it makes sense to start with the evacuation API integration from the start.

I did. This is an important feature for KubeVirt. There's a strong interest to have this feature merged sooner than later. Waiting until the Evacuation API is implemented/alpha would be the best course of actions. Yet, it's not a blocker for implementing the first alpha of this feature.

I know it is not a blocker. I see it more from the frame of:

Is this delivering enough value when implemented as a workaround to justify not waiting for the actual thing that you want to use? Is the feedback and initial users you will get now be worth implementing this, even if the thing will be quite different when the Evacuation API is here (in beta or when it is safe to use it or w/e)?

ingvagabund · 2024-06-18T08:43:22Z

Is this delivering enough value when implemented as a workaround to justify not waiting for the actual thing that you want to use? Is the feedback and initial users you will get now be worth implementing this, even if the thing will be quite different when the Evacuation API is here (in beta or when it is safe to use it or w/e)?

Disclaimer: this is my personal comment and it does not have to reflect the current reality.

We don't have any participating initial users besides the discussions I have had with some KubeVirt representatives. So currently there's only one user. The feedback is currently limited and influenced by the evacuation API proposal. I don't have any solid data to support anything I have proposed in the current proposal. The design was crafted in a best-effort manner in my best conscience. Taking into account many what-if scenarios. We still don't have many participating contributors in the descheduler community to extend the discussions/reviews and hand craft a better design. Practically, any contribution from my side is done either in my free time of when I prioritize my work in my current company.

I am aligned with your perspective and acknowledge you asked the right and important questions. I see two options. To wait (hard to say for how long) and collect more feedback with an assumption the design will improve. Or, pin the current design, produce v1alpha1 implementation, collect the feedback and see how well/bad the design proves the be.

knelasevero · 2024-06-19T16:53:47Z

We still don't have many participating contributors in the descheduler community to extend the discussions/reviews and hand craft a better design

You focus on the design reviews not being comprehensive for lack of contributors, but that is actually an argument to wait, no? Since if you consider the repo to have less support than it needs it would be harder to maintain a workaround than it would be to maintain what you envision to be the final implementation (that would also start at alpha, but would not have big changes of core internals, while evolving initially).

To wait (hard to say for how long) and collect more feedback with an assumption the design will improve. Or, pin the current design, produce v1alpha1 implementation, collect the feedback and see how well/bad the design proves the be.

I also see these two options, and I see pros and cons coming from both of them. I think the workaround design is great, I wouldn't consider it to be bad. I think it was detailed and a lot has been considered. Just raised this discussion to be the devil's advocate and check other paths. This is an important feature for KubeVirt -> but how urgent, I think that's the question. Also do they know of the plan to migrate later to the evacuation API approach? Would they explicitly prefer not to wait, or did they just say they need it asap?

with an assumption the design will improve

But you also assume that, or the goal of migrating to the intended final implementation using the actual evacuation API (and ditching the previous one) would not be written down here in you EP.

fossedihelm · 2024-07-01T05:51:34Z

Reading the great discussion here! To me, the proposal looks good, and thank you for this effort!

This is an important feature for KubeVirt -> but how urgent, I think that's the question. Also, do they know of the plan to migrate later to the evacuation API approach? Would they explicitly prefer not to wait, or did they just say they need it asap?

Answering this question: Yes, this is an essential feature for KubeVirt.
Let's start from the beginning: live migration is a key feature of KubeVirt; it allows the workload not to be compromised. For this reason, it is chosen as the default behavior during an eviction, drain, etc...
KubeVirt is also shipped by default with Openshift, and currently, Descheduler and KubeVirt behavior are not compatible with each other. Without this workaround, a cluster admin would have to choose between having an active KubeVirt core feature and the Descheduler, and this would put them in a difficult situation that would be best avoided.
As expressed above this is a workaround that will be overridden with the evacuation API and KubeVirt will move to the evacuation API accordingly.
IOW, yes, KubeVirt side it would definitely be better to have this ASAP.
Thank you!

knelasevero · 2024-07-01T10:00:24Z

Reading the great discussion here! To me, the proposal looks good, and thank you for this effort!

This is an important feature for KubeVirt -> but how urgent, I think that's the question. Also, do they know of the plan to migrate later to the evacuation API approach? Would they explicitly prefer not to wait, or did they just say they need it asap?

Answering this question: Yes, this is an essential feature for KubeVirt. Let's start from the beginning: live migration is a key feature of KubeVirt; it allows the workload not to be compromised. For this reason, it is chosen as the default behavior during an eviction, drain, etc... KubeVirt is also shipped by default with Openshift, and currently, Descheduler and KubeVirt behavior are not compatible with each other. Without this workaround, a cluster admin would have to choose between having an active KubeVirt core feature and the Descheduler, and this would put them in a difficult situation that would be best avoided. As expressed above this is a workaround that will be overridden with the evacuation API and KubeVirt will move to the evacuation API accordingly. IOW, yes, KubeVirt side it would definitely be better to have this ASAP. Thank you!

Thanks! Having your direct input here is important. Thanks for chiming in!

ingvagabund · 2024-07-08T13:53:31Z

I am currently refactoring the current code base so at the end the changes required to implement v1alpha1 of this KEP are minimal. The current PoC is under https://github.com/ingvagabund/descheduler/tree/eviction-in-background-code-2024-07-02.

damemi

looks good to me, and the evacuation API does seem like a more graceful evolution of the eviction API.

One question though, we have had requests for more immediate pod deletion (like #1253), and this seems to go a step further away from that. Should we have any continued support for direct eviction requests? Or is eviction going away entirely?

damemi · 2024-07-16T18:20:09Z

keps/1397-evictions-in-background/README.md

+### Annotation vs. evacuation API based eviction
+
+The [evacuation API](https://github.com/kubernetes/enhancements/pull/4565) is expected
+to replace the [eviction API](https://kubernetes.io/docs/concepts/scheduling-eviction/api-eviction/).


is there any timeline on deprecation for the eviction API?

Not yet. It will strongly depends on when the evacuation API graduates.

damemi · 2024-07-16T18:26:38Z

keps/1397-evictions-in-background/README.md

+  **Answer**: The first implementation will not reset any evacuation request.
+  The descheduler will account for existing requests and update the internal counters accordingally.
+  In the future a mechanism for deleting too old evacuation requests can be introduced.
+  I.e. based on a new `--max-evacuation-request-ttl` option.


should this be a policy config field?

Yeah. A policy config field is versioned. With a feature gate mechanism alpha feature fields can be ignored.

damemi · 2024-07-16T18:30:42Z

/approve

k8s-ci-robot · 2024-07-16T18:30:49Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: damemi

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 files:

~~OWNERS~~ [damemi]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

ingvagabund · 2024-07-17T06:46:58Z

Thank you Mike for going through the proposal.

Should we have any continued support for direct eviction requests? Or is eviction going away entirely?

@atiratree will know better though based on my understand the eviction API will not go away. Instead, users will still have option to choose between the evacuation (delayed eviction) and the eviction APIs. Yet, with strong recommendation to prefer the evacuation API as it will serve as a cluster scoped coordinator for any eviction.

Wrt. to #1253 we will need to decide whether the descheduler will eventually switch to the evacuation API by default or not. Yet, we will still keep the direct eviction API and any potential immediate deletion configurable. It will take some time before the evacuation API becomes GA.

atiratree · 2024-07-24T17:02:28Z

@atiratree will know better though based on my understand the eviction API will not go away. Instead, users will still have option to choose between the evacuation (delayed eviction) and the eviction APIs. Yet, with strong recommendation to prefer the evacuation API as it will serve as a cluster scoped coordinator for any eviction.

@ingvagabund yes the eviction API is not going away. A better version of it (e.g. Evacuation API, etc.) is still being discussed upstream in sig-apps and should eventually be used by all the major components like the descheduler, cluster autoscaler, etc. Normal users do not really use eviction much.

atiratree · 2024-07-24T17:04:26Z

in accordance with the above consensus
/lgtm

k8s-ci-robot · 2024-07-24T17:04:30Z

@atiratree: changing LGTM is restricted to collaborators

In response to this:

in accordance with the above consensus
/lgtm

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

k8s-ci-robot added cncf-cla: yes Indicates the PR's author has signed the CNCF CLA. size/XL Denotes a PR that changes 500-999 lines, ignoring generated files. labels Feb 26, 2024

k8s-ci-robot requested review from damemi and JaneLiuL February 26, 2024 09:59

ingvagabund force-pushed the evictions-in-background branch 2 times, most recently from 320169c to 5af0e72 Compare February 26, 2024 10:12

ingvagabund mentioned this pull request Feb 26, 2024

Design proposal: Descheduler support kubevirt/community#258

Merged

8 tasks

ingvagabund force-pushed the evictions-in-background branch 5 times, most recently from a53c30d to ec699a1 Compare May 7, 2024 13:43

ingvagabund changed the title ~~KEP: Extend descheduler eviction handling for eviction requests~~ KEP-1397: descheduler integration with evacuation API as an alternative to eviction API May 11, 2024

ingvagabund force-pushed the evictions-in-background branch 5 times, most recently from f1649b7 to c5cae17 Compare May 11, 2024 18:16

songtao98 reviewed May 14, 2024

View reviewed changes

ingvagabund force-pushed the evictions-in-background branch 3 times, most recently from 6fccd48 to 27acbd2 Compare May 14, 2024 11:34

k8s-ci-robot added size/XXL Denotes a PR that changes 1000+ lines, ignoring generated files. and removed size/XL Denotes a PR that changes 500-999 lines, ignoring generated files. labels May 14, 2024

KEP-1397: descheduler integration with evacuation API as an alternati…

b1c6e24

…ve to eviction API

knelasevero reviewed Jun 5, 2024

View reviewed changes

atiratree mentioned this pull request Jun 7, 2024

KEP-4563: Evacuation API kubernetes/enhancements#4565

Open

atiratree reviewed Jun 12, 2024

View reviewed changes

damemi reviewed Jul 16, 2024

View reviewed changes

k8s-ci-robot added the approved Indicates a PR has been approved by an approver from all required OWNERS files. label Jul 16, 2024

ingvagabund mentioned this pull request Jul 24, 2024

Introduce RequestEviction feature for evicting pods in background (KEP-1397) #1466

Merged

5 tasks

ingvagabund added the lgtm "Looks good to me", indicates that a PR is ready to be merged. label Jul 24, 2024

k8s-ci-robot merged commit f3569b5 into kubernetes-sigs:master Jul 24, 2024
9 checks passed

ingvagabund deleted the evictions-in-background branch July 24, 2024 18:47

KEP-1397: descheduler integration with evacuation API as an alternative to eviction API #1354

KEP-1397: descheduler integration with evacuation API as an alternative to eviction API #1354

Conversation

ingvagabund commented Feb 26, 2024 • edited Loading

songtao98 commented May 11, 2024 • edited Loading

ingvagabund commented May 11, 2024

songtao98 commented May 13, 2024

ingvagabund commented May 13, 2024

songtao98 commented May 14, 2024

Choose a reason for hiding this comment

knelasevero commented Jun 5, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ingvagabund Jun 18, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ingvagabund Jun 18, 2024 • edited Loading

Choose a reason for hiding this comment

knelasevero commented Jun 5, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ingvagabund Jun 18, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ingvagabund Jun 18, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ingvagabund Jun 18, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

atiratree commented Jun 12, 2024

ingvagabund commented Jun 17, 2024

ingvagabund commented Jun 17, 2024

knelasevero commented Jun 17, 2024

ingvagabund commented Jun 18, 2024 • edited Loading

knelasevero commented Jun 19, 2024 • edited Loading

fossedihelm commented Jul 1, 2024

knelasevero commented Jul 1, 2024

ingvagabund commented Jul 8, 2024

damemi left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

damemi commented Jul 16, 2024

k8s-ci-robot commented Jul 16, 2024

ingvagabund commented Jul 17, 2024

atiratree commented Jul 24, 2024

atiratree commented Jul 24, 2024

k8s-ci-robot commented Jul 24, 2024

ingvagabund commented Feb 26, 2024 •

edited

Loading

songtao98 commented May 11, 2024 •

edited

Loading

ingvagabund Jun 18, 2024 •

edited

Loading

ingvagabund Jun 18, 2024 •

edited

Loading

ingvagabund Jun 18, 2024 •

edited

Loading

ingvagabund Jun 18, 2024 •

edited

Loading

ingvagabund Jun 18, 2024 •

edited

Loading

ingvagabund commented Jun 18, 2024 •

edited

Loading

knelasevero commented Jun 19, 2024 •

edited

Loading