remove the PreCycle plugin from scheduler

[nirrozenbaum]

as discussed in scheduler design latest discussions, the plan is to completely remove the Pre/Post Cycle (previously called Pre/Post Schedule) from Scheduler. atm the only plugin that uses both is Prefix.
This PR completely removes PreCycle plugin from the scheduler.

As a side effect, this PR also fixes the following bug:

before this PR, the logic that calculated hashes in Prefix plugin PreCycle could have been buggy, since pods that have the longest prefix may be filtered out in the filter phase.

This PR is fixing this problem by calculating longest prefix only for the pods that passed the filtering phase.

cc: @ahg-g @kfswain @liu-cong

[netlify]

✅ Deploy Preview for gateway-api-inference-extension ready!

Name	Link
🔨 Latest commit	27548bf
🔍 Latest deploy log	https://app.netlify.com/projects/gateway-api-inference-extension/deploys/6836a20a4893d50008c51747
😎 Deploy Preview	https://deploy-preview-876--gateway-api-inference-extension.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[liu-cong]

/assign

[liu-cong]

the plan is to completely remove the Pre/Post Cycle (previously called Pre/Post Schedule) from Scheduler

I am fine with removing the preCycle as there is no use case for it anyway.

Is the plan to replace PostCycle with PostResponse? IMO we need something like PreRequest. Consider this scenario, two requests longPrefix+suffix1 and longPrefix+suffix2 arrive around the same time. Suppose each request will take a long time to get the response back, with PostResponse you can schedule them onto 2 different pods (no prefix affinity) due to the latency of getting the response back. With PreRequest we can still achieve prefix cache affinity.

[liu-cong]

/lgtm

[nirrozenbaum]

@liu-cong applied the suggested changes, need lgtm again :)

[liu-cong]

/lgtm

[nirrozenbaum]

the plan is to completely remove the Pre/Post Cycle (previously called Pre/Post Schedule) from Scheduler

I am fine with removing the preCycle as there is no use case for it anyway.

Is the plan to replace PostCycle with PostResponse? IMO we need something like PreRequest. Consider this scenario, two requests longPrefix+suffix1 and longPrefix+suffix2 arrive around the same time. Suppose each request will take a long time to get the response back, with PostResponse you can schedule them onto 2 different pods (no prefix affinity) due to the latency of getting the response back. With PreRequest we can still achieve prefix cache affinity.

@liu-cong the discussion on scheduler extension points and interfaces is still ongoing (in PR #845).
as far as I understand, the below diagram is where we're currently at in terms of the discussions (or something along these lines, e.g., names may be different). things can still change as this is still WIP.

[ahg-g]

/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: ahg-g, nirrozenbaum

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 [ahg-g]

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