Skip to content

Not a logging tool just time series db # Installation - Normal way Docs

Promethues is a service account, remeber to dislable it’s shell

sudo useradd -M -r -s /bin/false prometheus

Structure

  • Uses PromeQl query language
  • It’s based on the time series db (saved on the main node)
    • Exporters pulls the data over the http form the nodes
  • Targets are dicovered via service discovery or static configuration >U need to hve Prometheus server on main node >And Node Exporter on the others >protemtheus Structure

Nodes

  • They are simply called instances(any endpoint u can scrape)
    • They can be setup via the jsonfiles
  • Jobs
    • collections of instances

Push vs pull

Aritcle

Push *Refers to the end point pushing data into promethues

instance*

!It’s not a prefered method! - If u must push, therse a Protemtheus Push Gateway - This is not a agregator or distributed coutnter it’s a metric cache - !Pushed metrics are never removed without mannual purge! #### Pull Refers to the end point presenting http inforamtion and then having the promethues isntance scrape the data into database - Natively checks weather the node is down (becouse it’s pulling data obviously) - They roate if the scrapes are deleted ##### Retention time By defualt it st

--storage.tsdb.retatnion.time 31\

Service discovery

Can be either File based or Http Based - File based via node exporter ### Service - deuflt adress http://localhost:9100/metrics - can be any ohter port - Http based - U can add any metric/data u want - Response must be 200 - Contnent-Type must be application/json - Body must be valid JSON >Service discovery

Tracing and spans

  • Trace the hole journey of the request actions as it moves through all nodes in distributed stysem
  • Spans operation or work taking place on a service , web server responding to an individual request or function

Logging

By fefualt the prometheus logs to the /var/log/prometheus

Slos Slis Slas

Service Level Objectives

Prometheus limitaion

  • LTS a 10 year cycle
  • Every ^ weeks new minor realease cycle begins Local storage
  • It’s not clustered or repliceated so needs to be treaded like a disk
  • Compaction happens in the background
    • Initial 2h block are eventualy commpacted into longer blocks in the background
    • 10% of data retatnion
    • 32 days which ever is smaller

Dashboarding Basics

Instead of showing everything at once spilt it based on purpose - No more the 5 graphs on a console - No more hten 5 lines on each graph - Avoid mire then 20-30 entires in the rig-hand-sid table

Full Prometheus Model

protemtheus