From 8de21045315b5425dc4aaf9c2658ddb3edec5db7 Mon Sep 17 00:00:00 2001 From: helen Date: Sun, 1 Sep 2024 21:00:24 +0800 Subject: [PATCH] fix issue 1418: Mutually Exclusive Pods Uniformly Distributed Scheduling Signed-off-by: helen --- ...e-pods-uniformly-distributed-scheduling.md | 59 +++++++++++++++++++ 1 file changed, 59 insertions(+) create mode 100644 designs/mutually-exclusive-pods-uniformly-distributed-scheduling.md diff --git a/designs/mutually-exclusive-pods-uniformly-distributed-scheduling.md b/designs/mutually-exclusive-pods-uniformly-distributed-scheduling.md new file mode 100644 index 0000000000..f0ca870448 --- /dev/null +++ b/designs/mutually-exclusive-pods-uniformly-distributed-scheduling.md @@ -0,0 +1,59 @@ +# Mutually Exclusive Pods Uniformly Distributed Scheduling + +## Background + +> [issues 1418](https://github.com/kubernetes-sigs/karpenter/issues/1418) + +I have 10 pending pods: + +- **pod1**: 1c1g requests, with anti-affinity; cannot be scheduled on the same node as pod10 and pod9. +- **pod2 ~ pod8**: 1c1g requests; no anti-affinity is configured. +- **pod9**: 1c1g requests, with anti-affinity; cannot be scheduled on the same node as pod1 and pod10. +- **pod10**: 1c1g requests, with anti-affinity; cannot be scheduled on the same node as pod1 and pod9. + +In Karpenter, it creates three nodes, with the first node being super large and the remaining two being very small: + +- **node1**: c7a.4xlarge, 16c32g (8Pods) +- **node2**: c7a.large, 2c4g (1Pod) +- **node3**: c7a.large, 2c4g (1Pod) + +**Expected Behavior**: +I want the resources of the three nodes to be evenly distributed, like: + +- **node1**: c7a.4xlarge, 8c16g (4Pod) +- **node2**: c7a.xlarge, 4c8g (3Pod) +- **node3**: c7a.xlarge, 4c8g (3Pod) + +In this situation, the cluster will be more stable (e.g., draining one node will not cause most pods to be rescheduled). + +## Solutions + +By sorting in byCPUAndMemoryDescending, pods with the same specifications (cpu or memory request equal) and mutually exclusive pods (with PodAntiAffinity or TopologySpreadConstraints) are prioritized to be scheduled first. + +Code implementation: [pull 1548](https://github.com/kubernetes-sigs/karpenter/pull/1548) + +Pros/Cons + +- 👍👍 Simple implementation effort + +### Test Results + +```bash +❯ kubectl get nodeclaims +NAME TYPE CAPACITY ZONE NODE READY AGE +default-8wq87 c-8x-amd64-linux spot test-zone-d blissful-goldwasser-3014441860 True 67s +default-chvld c-4x-amd64-linux spot test-zone-b exciting-wescoff-4170611030 True 67s +default-kbr7n c-2x-amd64-linux spot test-zone-d vibrant-aryabhata-969189106 True 67s +❯ kubectl get pod -owide +NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES +nginx1-67877d4f4d-nbmj7 1/1 Running 0 77s 10.244.1.0 vibrant-aryabhata-969189106 +nginx10-6685645984-sjftg 1/1 Running 0 76s 10.244.2.2 exciting-wescoff-4170611030 +nginx2-5f45bfcb5b-flrlw 1/1 Running 0 77s 10.244.2.0 exciting-wescoff-4170611030 +nginx3-6b5495bfff-xt7d9 1/1 Running 0 77s 10.244.2.1 exciting-wescoff-4170611030 +nginx4-7bdd687bb6-nzc8f 1/1 Running 0 77s 10.244.3.5 blissful-goldwasser-3014441860 +nginx5-6b5d886fc7-6m57l 1/1 Running 0 77s 10.244.3.0 blissful-goldwasser-3014441860 +nginx6-bd5d6b9fb-x6lkq 1/1 Running 0 77s 10.244.3.2 blissful-goldwasser-3014441860 +nginx7-5559545b9f-xs5sm 1/1 Running 0 77s 10.244.3.4 blissful-goldwasser-3014441860 +nginx8-66bb679c4-zndwz 1/1 Running 0 76s 10.244.3.1 blissful-goldwasser-3014441860 +nginx9-6c47b869dd-nfds6 1/1 Running 0 76s 10.244.3.3 blissful-goldwasser-3014441860 +```