GCP Labels¶

GCE¶

Mandatory Labels requested by finance (enforced by policy)¶

tu= [agp, ods]
product = [offerwall, fairbid, direct, growth, exchange, brand_bidder, legacy_adcolony, appreciate]

env = [dev, prod, qa]¶

cost-type = [fixed, variable] (this labels defines fixed vs variable costs. Fo example, mysql we scale very rarely and it can be considered it as static. LoadBalancers are  dynamic type as it depends on the amount of traffic)

Our Labels  (enforced by policy)¶

owner = [berlin-devops, adcolony-devops, israel-devops]
sub-product = [ace, engine, opd, fairbid-backend, fairbid-frontend, exchange, etc.]
component = [redis, druid, vm-mysql, cloudsql-mysql, cloudsql-postgres, memstore-redis etc., gke]

sub-component - optional

role = [default, main, standby, secondary, name of job, etc ]

vm-type = [standalone, gke, mig, uig, dataproc, databricks, any services that creates vms]

lifecycle=[spot,normal] (gke nodes will get this label automatically)

Labels for tracking costs in gke (optional)¶

gke-node-pool-type = [static, nap] (we need to understand the value of this label)

Example of labels for calzone-aggregation spark job:Common mandatory labels:

sub-product = calzone-aggregation

component = spark

role = driver vs executor

vm-type = gke

Optional for gke:

Group for spot:

gke-node-pool-name = calzone-aggregation
gke-node-pool-lifecycle = spot
gke-node-pool-type = nap

Group for fallback to on-demand:
gke-node-pool-name = calzone-aggregation
gke-node-pool-lifecycle = od
gke-node-pool-type = nap

GCS Buckets¶

Mandatory Labels requested by finance (enforced by policy) -  possible to set only on bucket level and not on object or path!:
tu= [agp, ods]
product = [offerwall, fairbid, direct, growth, marketplace, adcolony]
owner = [berlin-devops, adcolony-devops, israel-devops]
data-owner = [, fairbid-backend, dtx-backend, ods...]
env = [dev, prod, qa]

accessibility = [private, public]

pii = [true, false]

Qustions:
*  How will we estimate costs for specific path in data lake buckets?