Define requests & limits for all placeos services

[viv-4]

• [x] Find & define reasonable defaults for dev/poc
• [x] Find & define reasonable defaults for prod
• [x] Guaranteed class for core & frontends (limits == requests)
• [x] Burstable class with limits for api, auth, staff (limits > requests)
• [x] Burstable class without limits for rubbersoul, triggers, dispatch, source (only requests)

notes:

• when there is a resource shortage on a node, BestEffort pods are killed first, then Burstable
• if a Burstable pod exceeds it's memory limit the kubernetes scheduler will stop/recreate it with OOMKilled status
• a pods requests must fit in a nodes available Allocatable Capacity to be scheduled to it

[viv-4]

Currently monitoring following resource definitions on 3 node cluster - 2cpu 8gb nodes, api, auth, core, staff - 2 pods | elastic 2 data, 3 master pod cluster | rethinkdb 3 pod cluster lens metrics (prometheus etc.) deployed Fairly low driver usage cluster - More active cluster required more overall resources and higher request/limit values on api & core memory

• core

          limits:
            cpu: 1
            memory: 1536Mi
          requests:
            cpu: 1
            memory: 1536Mi

• api

          limits:
            cpu: 100m
            memory: 100Mi
          requests:
            cpu: 10m
            memory: 50Mi

• auth

          limits:
            cpu: 500m
            memory: 500Mi
          requests:
            cpu: 10m
            memory: 400Mi

• staff

          limits:
            cpu: 25m
            memory: 250Mi
          requests:
            cpu: 10m
            memory: 50Mi

• frontends - frontends

          limits:
            cpu: 500m
            memory: 100Mi
          requests:
            cpu: 500m
            memory: 100Mi

• frontends - nginx

          limits:
            cpu: 50m
            memory: 100Mi
          requests:
            cpu: 50m
            memory: 100Mi

• search-ingest

          requests:
            cpu: 10m
            memory: 50Mi

• triggers

          requests:
            cpu: 10m
            memory: 50Mi

• dispatch (no active deployment to reference - 5Mi covers it running without activity)

          requests:
            cpu: 10m
            memory: 5Mi

• source

          requests:
            cpu: 10m
            memory: 50Mi

third-party charts using their defaults except elastic-data:

requests:
  cpu: 25m
  memory: 1536Mi

These requests are suitable for dev/poc cluster with pod redundancy and low usage

[viv-4]

Above committed to: https://github.com/place-labs/k8s-helm/commit/100eac454e567d2daa30ec0dfec26ca292505408 To be merged to master Prod values to go in new prod branch