Showing posts with label timeseries. Show all posts
Showing posts with label timeseries. Show all posts

2022-09-08

Getting started with InfluxDB

Today we're gonna learn about InfluxDB, a time series database that already been a standard for open source metric storage, a part of TICK stack (telegraf -- a metric collector like datadog-agent/fluentd, influx -- the database, chronograf -- visualizer like grafana, kapacitor -- an alert manager like prometheus alert-manager also for ETL). To install influx use this steps from this link (there's also docker). What's the cons of influx compared to other solution? Datadog is (very very) expensive, Prometheus seems too kube-oriented, Influx open source doesn't support HA and clustering, that's why I would rather use Clickhouse for time series / log collection (as data sink) and aggregate those logs as metrics (materialized view) and copy periodically to faster database like Tarantool. Influx have 2 query syntax, one is SQL-like called InfluxQL and other one Javascript+FP-like called Flux (it has |> pipe like most FP). Here's the example of docker and the query language:

docker run -d -e INFLUXDB_ADMIN_USER:user1 -e INFLUXDB_ADMIN_PASSWORD:pass1 --name influxdb1 influxdb

docker exec -it influxdb1 influx -host 127.0.0.1 -port 8086 -username user1 -password pass1

show databases
create database db1
use db1

-- show all tables (=measurements)

show measurements

-- tag/index always string, field can be float, int, bool, etc

insert into table1,tag1=a,tag2=b field1=1,field2=2

-- if not set, default time is in nanosecond of insert time

select * from "table1"
select * from table1 where tag1='a'

-- describe columns from all table:

show tag keys
show field keys from "table1"

-- select distinct tag from table1
show tag values from "table1" with key in ("tag1")

-- automatically delete after N days, default 0 = never delete
-- shard by N hours, default 168 hours (a week)
show retention policies

-- like partition in other database
show shards
show shard groups

I guess that's it what you need to know to get started with InfluxDB. If you're looking one that comparable with Clickhouse, you can check TDEngine or TiFlash.