Project-HAMi / HAMi

Heterogeneous AI Computing Virtualization Middleware
http://project-hami.io
Apache License 2.0
1.25k stars 253 forks source link

scheduler: client-side throttling, not priority and fairness #745

Open chaunceyjiang opened 1 month ago

chaunceyjiang commented 1 month ago

What happened:

│ I1227 02:28:17.812399       1 event.go:307] "Event occurred" object="u-dad49dd32cdf/test-222-76c478b9c5-sq794" fieldPath="" kind="Pod" apiVersion="v1" type="Warning" reason="FilteringFailed" message="no available node, all node scores do not meet"                                                                  │
│ I1227 02:28:17.977042       1 request.go:629] Waited for 197.657513ms due to client-side throttling, not priority and fairness, request: PATCH:https://10.233.0.1:443/api/v1/namespaces/u-8c6ddc72f5a0/events/qwen-0-5-7fc4c6c8f9-h5tvm.1814e77ce9f9cd1d                                                                 │
│ I1227 02:28:18.176980       1 request.go:629] Waited for 196.317212ms due to client-side throttling, not priority and fairness, request: PATCH:https://10.233.0.1:443/api/v1/namespaces/u-8c6ddc72f5a0/events/qwen-0-5-7fc4c6c8f9-rwf2g.1814e77ceb07cb5c                                                                 │
│ I1227 02:28:18.376885       1 request.go:629] Waited for 196.355669ms due to client-side throttling, not priority and fairness, request: PATCH:https://10.233.0.1:443/api/v1/namespaces/u-8c6ddc72f5a0/events/qwen-0-5-7fc4c6c8f9-nr4lq.1814e77ceb90fa99                                                                 │
│ I1227 02:28:18.576398       1 request.go:629] Waited for 195.351409ms due to client-side throttling, not priority and fairness, request: PATCH:https://10.233.0.1:443/api/v1/namespaces/u-8c6ddc72f5a0/events/qwen-0-5-7fc4c6c8f9-csw5w.1814e77cec592297                                                                 │
│ I1227 02:28:18.776450       1 request.go:629] Waited for 196.339784ms due to client-side throttling, not priority and fairness, request: PATCH:https://10.233.0.1:443/api/v1/namespaces/u-8c6ddc72f5a0/events/qwen-0-5-7fc4c6c8f9-9ttks.1814e77ceccba411

What you expected to happen:

When a large number of pods are discovered for scheduling, client-side throttling is triggered.

How to reproduce it (as minimally and precisely as possible):

Anything else we need to know?:

Environment:

chaunceyjiang commented 1 month ago

/assgin