Skip to content

feat: Moved the Routing Sidecar from its own repo to the inference-scheduler repo#379

Merged
shmuelk merged 10 commits intollm-d:mainfrom
shmuelk:sidecar
Oct 23, 2025
Merged

feat: Moved the Routing Sidecar from its own repo to the inference-scheduler repo#379
shmuelk merged 10 commits intollm-d:mainfrom
shmuelk:sidecar

Conversation

@shmuelk
Copy link
Collaborator

@shmuelk shmuelk commented Oct 19, 2025

This PR moves the Routing Sidecar from its own repo into the inference-scheduler repo.

This includes:

  • The code
  • Makefile targets
  • Unit tests
  • Dockerfile
  • CI steps

The Routing Sidecar End to End tests will be added in a followup PR

Refs: #335

@nirrozenbaum
Copy link
Collaborator

@shmuelk I think we need Lionel for this review.

@shmuelk shmuelk requested a review from lionelvillard October 19, 2025 10:06
@elevran elevran moved this to In review in llm-d-inference-scheduler Oct 19, 2025
Comment on lines 52 to 53
// ConnectorNIXLV1 enables the (now deprecated) P/D NIXL v1 protocol
ConnectorNIXLV1 = "nixl"

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is maybe a good time to get rid of deprecated protocols. AFAIK, only nixlv2 is used.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'll remove nixl V1. There is talk about LMCache....

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done

See the License for the specific language governing permissions and
limitations under the License.
*/
package main

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you llm-d-routing-sidecar to something shorted? I would recommend proxy as this is a well-known term in the P/D context. Or pd_proxy if you want the context to be explicit.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Renamed to pd-sidecar

limitations under the License.
*/

package signals

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done.

Copy link

@lionelvillard lionelvillard left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Added few minor comments.

@shmuelk shmuelk force-pushed the sidecar branch 2 times, most recently from 0658063 to 24ac8b9 Compare October 22, 2025 16:19
Copy link
Collaborator

@elevran elevran left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

please open an issue to change makefile targets and variables so that EPP and sidecar are treated the same.

shell: bash
run: |
make test
make test sidecar-test
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: suggest using consistent and qualified target names across scheduler and sidecar and retaining the unqualified targets for cross exe chores.
For example: test calls test-sidecar and test-epp. Same for build.
Same naming can be applied to docker files (e.g. dockerfile-sidecar) and that means we can use % targets for commanlity.

Not critical for this PR, perhaps open an issue?

TARGETOS ?= $(shell go env GOOS)
TARGETARCH ?= $(shell go env GOARCH)
PROJECT_NAME ?= llm-d-inference-scheduler
SIDECAR_IMAGE_NAME ?= llm-d-routing-sidecar
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: ditto for naming conventions of variables in the Makfile (EPP_ and SIDECAR_...)

@elevran
Copy link
Collaborator

elevran commented Oct 23, 2025

/lgtm
/approve

@github-actions github-actions bot added the lgtm "Looks good to me", indicates that a PR is ready to be merged. label Oct 23, 2025
@elevran
Copy link
Collaborator

elevran commented Oct 23, 2025

@shmuelk please sign all commits before PR can be merged. Currently most are missing a verified signature

Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>
Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>
Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>
Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>
Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>
Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>
Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>
Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>
Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>
Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>
@elevran
Copy link
Collaborator

elevran commented Oct 23, 2025

/lgtm
/approve

@shmuelk shmuelk merged commit 6e2bb8a into llm-d:main Oct 23, 2025
6 checks passed
@github-project-automation github-project-automation bot moved this from In review to Done in llm-d-inference-scheduler Oct 23, 2025
@shmuelk shmuelk deleted the sidecar branch October 23, 2025 12:55
guygir pushed a commit to guygir/llm-d-inference-scheduler that referenced this pull request Nov 2, 2025
…heduler repo (llm-d#379)

* Moved prefill header definition to common import

Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>

* Moved Routing Sidecar into this repo

Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>

* Moved Routing Sidecar tests into this repo

Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>

* Moved Routing Sidecar Dockerfile into this repo

Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>

* Added Routing Sidecar to Makefile

Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>

* Added Routing Sidecar to CI stream

Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>

* Fixed lint error

Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>

* Review fixes and added version info

Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>

* Test Nixl V2 instead of the deleted Nixl V1

Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>

* Fixed lint errors

Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>

---------

Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

lgtm "Looks good to me", indicates that a PR is ready to be merged.

Projects

Archived in project

Development

Successfully merging this pull request may close these issues.

4 participants