https://signoz.io logo
#general
Title
# general
a

Agustin Lucchetti

08/10/2023, 3:47 PM
H! i’m trying to deploy otel+signoz on a large fleet of EC2, managed by multiple autoscaling groups. What’s the recommend architecture pattern for it? 1. Each EC2 instance has an otel collector that fowards the data to signoz otel-collector 2. Each EC2 instance has an otel collector that fowards the data directly to the clickhouse DB 3. The signoz otel-collector scraps the data from each EC2 instance’s otel collector 4. The signoz otel-collector scraps the data directly form eachh EC2 instance receiver (prometheus, jaeger, etc) Im guessing 3 and 4 would require some sort of service discovery
s

Srikanth Chekuri

08/10/2023, 3:59 PM
I would suggest going with option one.
a

Alexei Zenin

08/10/2023, 6:41 PM
Option 1 is how I do it in production with billions of rows in clickhouse and scales well. Can run the otel-collector as a daemon service if using ECS
More docs on the OTEL site here around deployment where you can gauge the tradeoffs on how many layers of OTEL collectors you want before reaching your “backend” storage for the metrics/traces/logs: https://opentelemetry.io/docs/collector/deployment/
a

Agustin Lucchetti

08/10/2023, 6:51 PM
thanks folks!
i will take a look at those docs
3 Views