I’m trying to estimate the cost of datadog when i instrument temporal with it. Generally, datadog charges by series, and each series = metric name + tag.
Therefore, every permutation of tag + metric name = 1 series.
temporal/defs.go at master · temporalio/temporal · GitHub → Seeing this list of metric names, what’s the combination of possible tags for each of these?
Consequentially, roughly how many series could I expect? Thanks!
I believe tag=label in this context.
Each metric name is reported with one set of labels. However each label can have multiple potential values.
Roughly speaking, most metrics have within 7 different labels plus any labels you additionally specify via config.
Label values can differ greatly depending on config, usage and label.
Ex.:
Client latency has namespace, operation, service_role labels. However namespace list is dependent on amount of namespaces you register. Operation values are predefined and are a limited list, differ per metric (defs.go can be used as a reference). service_role has 6 possible values atm.
client_latency_sum {namespace=“all”,operation=“HistoryClientStartWorkflowExecution”,service_role=“history”}