Preliminary study on temporal database incluxdb 2.2

What is a time series database ？ There is no popular science here , Please Baidu . Timing data is a scenario of writing more and reading less .

InfluxDB use Go Language writing , Open source , It should be good . But the disadvantage is ： The stand-alone version is free and open source , There is a charge for the cluster version .

install

Download the database separately Server And command line tools CLI, Two separate programs . After installation , perform influx start-up Server, Pay attention to exposure 8086 Default port .

influx

This is a temporary start , We changed the daemon to execute ,

nohup ./influx &

Then on Server Configure accordingly , Yes WebUI The interface of , Visit its 8086 Web The service can be , As shown in the figure below .

Set the initial user , All required items .

Successful initialization .

Java Client calls

Our application is Java, So do it ：

<dependency>
    <groupId>com.influxdb</groupId>
    <artifactId>influxdb-client-java</artifactId>
    <version>6.1.0</version>
</dependency>

Measurement

Excerpt from netizens ：

In a relational database , We are all on the watch table Query data in , According to the understanding of inertial thinking , Will also put influxdb Of measurement Understood as a table , Then when you query, you will naturally bring _measurement, However, in the actual query, this is only used as influxdb Query criteria for , In fact, it is an optional condition , Even if not, it can still be queried correctly . But when writing, you must measurement Of .

It can also be seen from here ,influxdb Of measurement With a relational database table Not exactly a concept , stay influxdb in measurement It's just bucket As a grouping .influxdb In fact, there is only one data structure , therefore measurement It's the only one ,measurement It will be automatically created when writing data , If the data does not exist ,measurement It will disappear naturally .

Simply compare the two ：

Write data

Writing tools can support ：influx Command line or API client , For example, we write Java It is officially provided Java-client. Writing mode can support Line protocol Data format or Java Entity Bean, recommend Java Bean intuitive .

asynchronous / Sync

Java Data written by the client can be divided into synchronous writing and asynchronous writing .

WriteApiBlocking writeApi = client.getWriteApiBlocking(); //  Blocking , Synchronization 
WriteApi makeWriteApi = client.makeWriteApi(); //  Non blocking , Asynchronous 

Better asynchronous performance , You can write to the database once in either of the following two cases ：

• Timer flush operation , Such as once a second
• Write data to 5000 pen , Write once , Control this quantity to batch_size（ Adjustable ）

Before writing data , Data is overstocked buffer In cache .

//  write in , Specify a precision of  ns  nanosecond 
writeApi.writeMeasurement(WritePrecision.NS, vo);

Browsing data

Database tools we will think NativeCat And so on. , but influx What about ？ Please rest assured that it is officially provided UI Tools , And it's beautifully made , Access the deployment location with a browser , Such as http://localhost:8086, Click on Data Explorer You can browse .

Flux

Flux（flux: n. Traffic ; Changes ; unstable ; Flow out ; vt. Melt ; Treat with flux ; vi. melt ; Flow out ） yes InfluxDB 2.0 A query language introduced , The new version v2 Abandoned v1 Similar SQL grammar , Completely use self-made query methods , be called "Flux". Every Flux Queries need to include the following parts ：1. data source ,2 Time range ,3 Data filters .

data source ：bucket Identify the name of the database

from(bucket:"example-bucket")

Time range ,stop It's not necessary , The time range can be a specific time （UTC Time ） Or timestamps , It can also be a relative time range , Such as -1h Show the past 1 Data in hours （ Relative to the current time ）, Optional units are s,m( minute ),h,d,mo( month ),y

  |> range(start: -1h, stop: -10m)

When querying time series data ,Flux Need a time frame ." unbounded " Queries are very resource intensive , As a protective measure ,Flux The database will not be queried without a specified range .

Data filters , Multiple filters can be used and or or Connect , Or start another filter

|> filter(fn: (r) =>)

filter The optional values for are ：_measurement ,_field ,_value,_time, Some tag The name of

Generate query data （ Optional ）

|> yield()

The output table usually contains ：_start, _stop, _field,_value, _measurement,_time,[tag name ] Field

Every flux Grammar is written in “from” Start , Every other part needs to be marked with " |> " start .

About Tag Field

Start using Flux when , Find even simple SQL where The specified condition query cannot be done , Very disappointed ！ I asked a lot and didn't know , Later, I read the English documents tag Talent inspiration , There is no designation tag Why .influx in , Distinguish ordinary field and tag field, The former is not indexed , So it can't be searched , The latter can . So when you want to search a field , It must be specified as tag field, Like the one below uavId Field

/** *  The battery 、 voltage  */
@Measurement(name = "Power")
public class Power extends InfluxValueObject {
    
	@Column(tag = true)
	public String uavId;

	@Column
	public Integer batteryRemaining;

	@Column
	public Float voltageBattery;
}

If you build too many indexes , write in 、 Query performance will decline .

Two kinds of function

The first one is ： aggregate function; The second kind ：selector function. The most important difference between these two types of functions is ,aggregate funciton Is to return a data record through aggregation ;selector funciton Is to return a set of original data . These two kinds of functions can be mixed in some places , Sometimes you can't .

similar SQL Of OR perhaps IN Inquire about

• stay filter Use in or Connect ： |> filter(fn: (r) => r.eq == "1" or r.eq == "2")
• Use contains function ： stay from Previously defined array fields = ["1", "2"], And then in filter Use in |> filter(fn: (r) => contains(value: r.eq, set: fields))

The data backup

Reference resources ：https://www.sunzhongwei.com/influxdb-20-data-backup-recovery-exportimport?from=bottom