changeset 144:476b23425f8c

jobs allowed on dash, but not really busy ones
author drewp@bigasterisk.com
date Sun, 19 Feb 2023 21:35:22 -0800
parents 6b4226845254
children ee0384eebee5
files k8s_reserve/deploy.yaml k8s_reserve/readme
diffstat 2 files changed, 30 insertions(+), 0 deletions(-) [+]
line wrap: on
line diff
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/k8s_reserve/deploy.yaml	Sun Feb 19 21:35:22 2023 -0800
@@ -0,0 +1,27 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: reserve-cpu-dash
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app.kubernetes.io/name: reserve-cpu-dash
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/name: reserve-cpu-dash
+      annotations: { prometheus.io/scrape: "false" }
+    spec:
+      containers:
+        - name: sleep
+          image: "docker.io/rancher/pause:3.6"
+          resources: {requests: {cpu: "3"}}
+      affinity:
+        nodeAffinity:
+          requiredDuringSchedulingIgnoredDuringExecution:
+            nodeSelectorTerms:
+            - matchExpressions:
+              - key: "kubernetes.io/hostname"
+                operator: In
+                values: ["dash"]
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/k8s_reserve/readme	Sun Feb 19 21:35:22 2023 -0800
@@ -0,0 +1,3 @@
+It's ok for k8s to use dash's ram, but it can also tie up so much cpu (notably with frigate) that I notice stalls on my desktop.
+
+The plan is to make a sleeper job with required.cpu=N to reserve N cpus from getting scheduled with real work.