Why monitor

Question 1

Name main reasons for monitoring

Answer 1

• detect issue fast, before client calls
• clear picture of resources usage - spending optimisation

Question 2

What to monitor

Answer 2

• resources usage: CPU, Memory, Disk space
• request rate
• error rate
• number of instances of apps

Question 2

name possible samples

Answer 2

• timestamp + value, (example: temp in New York given day),
• timeseries: list of samples ordered by time (example: temp in New York for few days in a row)

Question 3

What is TSDB?

Answer 3

• timeseries database (TSDB): special type of DB used to store timeseries, optimised toward querying by time

Question 4

explain quantiles and percentiles

Answer 4

they are the same but represented differently: for example 90th quantile is 0.9 percentile,
- 90% of the time a request takes no more than 1 s OR
- 0.9 quantile equals 1 OR
- 90th percentile equals 1

Question 5

give an example of metric name and it’s labels

Answer 5

temperature {city=’New York’, value=’33’}

temperature is metric name, city and value are labels

Question 6

what are jobs and instance labels?

Answer 6

Prometheus gives those two values out of the box. Job is a service name, for example location-service, instance is a node on which the service is running on

Question 7

what is instant vector?

Answer 7

list of samples where all of them contain the same timestamp, for example temperature in different places and different values but at the same time!

Question 7

what is range vector?

Answer 7

subset of samples from a timeseries in a given time range, for example temperature in LosAngeles from last 20s