May 2019
Intermediate to advanced
442 pages
11h 36m
English
A good starting point for scaling when a single Prometheus instance isn't enough is to split scrape jobs into logical groups and then assign these groups to different Prometheus instances. This is called vertical sharding. The groups can be made around anything that makes sense to you: by architecture/scope (frontend, backend, databases), by layers (OS-level metrics, infrastructure services, applications), by internal organization vertical, by team, by network security boundary (so that scrapes don't need to cross firewalls), or even by application cluster.
The following diagram exemplifies what this type of vertical sharding looks like:
This type of sharding also ...
Read now
Unlock full access