15b0b26a05
After node2 OOM incident, right-size memory across the cluster by setting requests=limits based on max_over_time(container_memory_working_set_bytes[7d]) with 1.3x headroom. Eliminates ~37Gi overcommit gap. Categories: - Safe equalization (50 containers): set req=lim where max7d well within target - Limit increases (8 containers): raise limits for services spiking above current - No Prometheus data (12 containers): conservatively set lim=req - Exception: nextcloud keeps req=256Mi/lim=8Gi due to Apache memory spikes Also increased dbaas namespace quota from 12Gi to 16Gi to accommodate mysql 4Gi limits across 3 replicas.
62 lines
1.3 KiB
HCL
62 lines
1.3 KiB
HCL
resource "helm_release" "caretta" {
|
|
namespace = kubernetes_namespace.monitoring.metadata[0].name
|
|
create_namespace = true
|
|
name = "caretta"
|
|
|
|
repository = "https://helm.groundcover.com/"
|
|
chart = "caretta"
|
|
version = "0.0.16"
|
|
|
|
values = [yamlencode({
|
|
grafana = {
|
|
enabled = false
|
|
}
|
|
victoria-metrics-single = {
|
|
enabled = false
|
|
}
|
|
tolerations = [
|
|
{
|
|
key = "node-role.kubernetes.io/control-plane"
|
|
operator = "Exists"
|
|
effect = "NoSchedule"
|
|
},
|
|
{
|
|
key = "nvidia.com/gpu"
|
|
operator = "Exists"
|
|
effect = "NoSchedule"
|
|
}
|
|
]
|
|
resources = {
|
|
requests = {
|
|
cpu = "10m"
|
|
memory = "768Mi"
|
|
}
|
|
limits = {
|
|
memory = "768Mi"
|
|
}
|
|
}
|
|
})]
|
|
}
|
|
|
|
resource "kubernetes_service" "caretta_metrics" {
|
|
metadata {
|
|
name = "caretta-metrics"
|
|
namespace = kubernetes_namespace.monitoring.metadata[0].name
|
|
labels = {
|
|
app = "caretta"
|
|
}
|
|
}
|
|
spec {
|
|
selector = {
|
|
app = "caretta"
|
|
}
|
|
port {
|
|
name = "metrics"
|
|
port = 7117
|
|
target_port = 7117
|
|
protocol = "TCP"
|
|
}
|
|
}
|
|
}
|
|
|
|
# Caretta dashboard is now loaded via the grafana_dashboards for_each in grafana.tf
|