Sample configuration for AWS EC2 handling 100k metrics/sec (large)

Download this sample AWS EC2 handling 100k metrics/sec (large) configuration provided by ITRS.

# Example Obcerv configuration for AWS EC2 handling 100k metrics/sec.
#
# Nodes:
# - (3) m5.8xlarge (32 CPU, 128GiB Memory) for Timescale
# - (5) c5.4xlarge (16 CPU, 32GiB Memory) for all other workloads
#
# The resource requests for Timescale total 48 cores and 360GiB memory.
# The resource requests for the other workloads total ~63 cores and ~113GiB memory.
# These totals include Linkerd resources.
#
# Disk requirements:
# - Timescale:
#   - 16 TiB data disk for each replica (x3)
#   - 75 GiB WAL disk for each replica (x3)
# - Kafka: 400 GiB for each replica (x3)
# - Loki: 30 GiB for each replica (x1)
# - Zookeeper: 1 GiB for each replica (x3)
# - etcd: 1 GiB for each replica (x3)
# - Downsampled Metrics:
#   - Raw: 5 GiB for each replica (x6)
#   - Bucketed: 5 GiB for each replica (x6)
#
# The configuration references a StorageClass named `io1-25` which uses io1 with 25 iopsPerGB - you can create
# this class or change the config to use a class of your own, but it should be similar in performance.
#
# This configuration is based upon a certain number of Obcerv entities, average metrics per entity, and
# average metrics collection interval. The following function can be used to figure out what type of load to expect:
#
# metrics/sec = (Obcerv entities * metrics/entity) / average metrics collection interval
#
# In this example configuration, we have the following:
#
# 100,000 metrics/sec = (250,000 Obcerv entities * 4 metrics/entity) / 10 seconds average metrics collection interval
#

defaultStorageClass: "gp2"
apps:
  externalHostname: "obcerv.mydomain.internal"
  ingress:
    annotations:
      kubernetes.io/ingress.class: "nginx"
      nginx.org/mergeable-ingress-type: "master"
ingestion:
  externalHostname: "obcerv-ingestion.mydomain.internal"
  replicas: 3
  ingress:
    annotations:
      kubernetes.io/ingress.class: "nginx"
      nginx.ingress.kubernetes.io/backend-protocol: "GRPC"
  resources:
    requests:
      memory: "512Mi"
      cpu: "500m"
    limits:
      memory: "768Mi"
      cpu: "1"
iam:
  ingress:
    annotations:
      kubernetes.io/ingress.class: "nginx"
      nginx.org/mergeable-ingress-type: "minion"
zookeeper:
  replicas: 3
  resources:
    requests:
      memory: "256Mi"
      cpu: "200m"
    limits:
      memory: "512Mi"
      cpu: "200m"
kafka:
  replicas: 3
  diskSize: "400Gi"
  storageClass: "io1-25"
  defaultPartitions: 12
  consumer:
    fetchMaxWaitMs: 2000
    fetchMinBytes: 8388608
  resources:
    requests:
      memory: "12Gi"
      cpu: "3"
    limits:
      memory: "12Gi"
      cpu: "4"
timescale:
  clusterSize: 3
  dataDiskSize: "100Gi"
  dataStorageClass: "io1-25"
  timeseriesDiskCount: 4
  timeseriesDiskSize: "4Ti" # Max disk size for AWS io1 is 16Ti
  timeseriesStorageClass: "io1-25"
  walDiskSize: "75Gi"
  walStorageClass: "io1-25"
  resources:
    requests:
      memory: "120Gi"
      cpu: "16"
    limits:
      memory: "120Gi"
      cpu: "16"
  compressAfter: 1h
  nodeSelector:
    instancegroup: timescale-nodes
  tolerations:
  - key: dedicated
    operator: Equal
    value: timescale-nodes
    effect: NoSchedule
  retention:
    entity_attributes:
      chunkSize: 2d
    metrics:
      chunkSize: 20m
      retention: 20d
    metrics_5m:
      chunkSize: 1h
      retention: 30d
    metrics_15m:
      chunkSize: 2h
      retention: 60d
    metrics_1h:
      chunkSize: 6h
      retention: 90d
    metrics_3h:
      chunkSize: 12h
      retention: 120d
    metrics_12h:
      chunkSize: 2d
      retention: 180d
    metrics_1d:
      chunkSize: 3d
      retention: 1y
    statuses:
      chunkSize: 7d
      retention: 1y
loki:
  diskSize: "30Gi"
  ingestionBurstSize: 12
  ingestionRateLimit: 8
  maxPayloadSize: 8388608
  resources:
    requests:
      memory: "1Gi"
      cpu: "500m"
    limits:
      memory: "1Gi"
      cpu: "1"
sinkd:
  replicas: 4
  rawReplicas: 4
  jvmOpts: "-Xms768M -Xmx768M -XX:MaxDirectMemorySize=100M"
  entityCacheMaxSize: 350000
  timeseriesCacheMaxSize: 700000
  resources:
    requests:
      memory: "1152Mi"
      cpu: "250m"
    limits:
      memory: "1152Mi"
      cpu: "400m"
  rawResources:
    requests:
      memory: "1Gi"
      cpu: "250m"
    limits:
      memory: "1Gi"
      cpu: "400m"
platformd:
  replicas: 2
  resources:
    requests:
      memory: "1536Mi"
      cpu: "1"
    limits:
      memory: "2Gi"
      cpu: "1500m"
dpd:
  replicas: 3
  jvmOpts: "-Xms3584M -Xmx3584M"
  resources:
    requests:
      memory: "3600Mi"
      cpu: "6"
    limits:
      memory: "4Gi"
      cpu: "8"
metricForecastd:
  resources:
    requests:
      memory: "512Mi"
      cpu: "250m"
    limits:
      memory: "768Mi"
      cpu: "500m"
downsampledMetricsStream:
  replicas: 3
  storageClass: "io1-25"
  bucketedReplicas: 4
  rawRocksdb:
    totalOffHeapMemory: 268435456
    indexFilterRatio: 0.25
    totalMemTableMemory: 201326592
    blockSize: 32768
    writeBufferSize: 33554432
  bucketedRocksdb:
    totalOffHeapMemory: 33554432
    indexFilterRatio: 0.25
    totalMemTableMemory: 25165824
    blockSize: 16384
    writeBufferSize: 8388608
  resources:
    requests:
      memory: "3Gi"
      cpu: "1"
    limits:
      memory: "3Gi"
      cpu: "1500m"
  bucketedResources:
    requests:
      memory: "4Gi"
      cpu: "1500m"
    limits:
      memory: "4Gi"
      cpu: "2"
entityStream:
  intermediate:
    replicas: 2
    resources:
      requests:
        memory: "1Gi"
        cpu: "1"
      limits:
        memory: "1536Mi"
        cpu: "1500m"
  final:
    resources:
      requests:
        memory: "512Mi"
        cpu: "1"
      limits:
        memory: "2500Mi"
        cpu: "1500m"
etcd:
  replicas: 3
collection:
  metrics:
    resources:
      requests:
        memory: "768Mi"
        cpu: "200m"
      limits:
        memory: "1Gi"
        cpu: "250m"
["Obcerv"] ["User Guide", "Technical Reference"]

Was this topic helpful?