Skip to content

The core of BBData: its API (and database definition)

Notifications You must be signed in to change notification settings

big-building-data/bbdata-api

Repository files navigation

BBData API

Build & Tests

This repository is the cornerstone of BBData. It contains:

  1. a SpringBoot Application exposing a REST API for submitting measures, administering objects, users, etc.
  2. dockerfiles and docker-compose for local testing, including: MySQL, Cassandra, Kafka
  3. the definition of the two databases at the center of BBData: MySQL & Cassandra

Development setup

Prerequisites

  • Java 1.8+
  • IntelliJ IDE with Kotlin support
  • Docker

Setup

Open the project in IntelliJ and let it work. Once finished, you should be able to simply run the app by launching the main class ch.derlin.bbdata.BBDataApplication (open it and right-click > run).

Of course, you will need MySQL, Cassandra and Kafka running for the whole API to run (to skip some of those deps, the the Profiles section).

Cassandra, MySQL and Kafka

To setup the three dependant services, have a look at the other folder. It contains all the files needed for a production setup, as well as instruction on how to run a Docker container for local dev/testing.

Profiles

By default, the app will launch with everything turned on, and will try to connect to MySQL, Cassandra and Kafka on localhost on default ports (see src/main/resources/application.properties).

Profiles let you disable some parts of the application. This is very useful for quick testing. To enable specific profiles, use the -Dspring.profiles.active=XX[,YY] JVM argument. On IntelliJ: Edit Configurations ... > VM Options.

Currently available profiles (see the class ch.derlin.bbdata.api.Profiles):

  • unsecured: all endpoints will be available without apikeys; the userId is automatically set to 1;
  • input: will only register the "input" endpoint (POST /objects/values);
  • output: will only register the "output" endpoints (everything BUT the one above);
  • noc: short for "No Cassandra". It won't register endpoints needing a Cassandra connection (input and values);
  • sqlstats: use MySQL to store objects statistics, instead of Cassandra BEWARE this is waaaay slower !!

Profiles can be combined (when it makes sense).

Examples:

Output only:

java -jar bbdata.jar -Dspring.profiles.active=output

Output only, no security checks:

java -jar bbdata.jar -Dspring.profiles.active=output,unsecured

Output only, no security checks and no cassandra (note that the output profile is not needed, as no Cassandra means no input):

java -jar bbdata.jar -Dspring.profiles.active=noc,unsecured

Hidden system variables

There are two system variables that can be set (export XX=YY), mostly for use in tests:

  • BB_NO_KAFKA=<bool>: this turns off the publication of augmented values to kafka in the input endpoint. Useful when we don't want to spawn a Kafka container.
  • UNSECURED_BBUSER=<int>: this determines which userId is used as default when the UNSECURED profile is turned off. Default to 1.

In test files, those variables can be set using the @SpringBootTest annotation, e.g.:

@SpringBootTest(webEnvironment = SpringBootTest.WebEnvironment.RANDOM_PORT, properties = ["UNSECURED_BBUSER=2"])

Production

To deploy in production, you need to build the jar and override some properties in order to connect to the correct services.

  1. to build the jar: ./gradlew bootJar, the jar will be created as ./build/libs/*.jar
  2. to specify properties from a file outside the jar: https://www.baeldung.com/spring-properties-file-outside-jar

Minimal properties to provide

Here is a sample file, values to change are in UPPERCASE:

## MySQL properties
spring.datasource.url = jdbc:mysql://HOST:PORT/bbdata2?autoReconnect=true&useUnicode=true&characterEncoding=UTF-8&allowMultiQueries=true&allowPublicKeyRetrieval=true&serverTimezone=UTC
spring.datasource.username=bbdata-admin
spring.datasource.password=PASSWORD

## Cassandra properties
spring.data.cassandra.contact-points=IP_1,IP_2,IP_X
spring.data.cassandra.consistency-level=quorum

## Kafka properties
spring.kafka.producer.bootstrap-servers=HOST:PORT
spring.kafka.template.default-topic=bbdata2-augmented

## Secured actuators endpoints (ensure the port is not available to the outside world)
management.server.port=SECURED-PORT

Executing the jar

The jar is fully executable, meaning you can do:

# default
./bbdata-api-*.jar

# with profiles: no "-D", but "--" instead !
./bbdata-api-*.jar --spring.profiles.active=unsecured,noc

Or, you can use the old school way:

# default
java -jar bbdata-api-*.jar

# with profiles: no -D !
java -Dspring.profiles.active=unsecured,noc -jar bbdata-api-*.jar

If you want to use a properties file that do not match the default names (e.g. development.properties), set the spring.config.additional-location (so it is loaded alongside and not instead of the default properties):

./bbdata-api-*.jar --spring.config.additional-location=development.properties

Caching

Caching is interesting in one part of the application, namely the input endpoint POST /values. The application needs to fetch the metadata associated to an object (and a token) on each request. To speed it up, the application can be configured to cache those metadata using either an in-memory concurrent hashmap, or an external database such as redis.

To enable/disable caching, set the spring.cache.type property. The former can take three values:

  • none: value by default, no caching at all,
  • simple: in-memory caching, only possible when the application runs in standalone mode, with both input and output enabled,
  • redis: use a redis instance for caching; this is mandatory if you want to deploy input and output separately and still use a cache.

Example:

./bbdata-api-*.jar --spring.cache.type=simple

Redis default properties are shown below (can be overriden from a property file). When using spring.cache.type=redis, you can override any of those properties to point to your redis server instance:

# Relevant properties in order to use redis (localhost and 6379 are the default)
spring.cache.type=redis
spring.redis.host=localhost
spring.redis.port=6379

By default, the cache logger is set to INFO. Feel free to change it using:

# default to INFO
logging.level.org.springframework.cache=TRACE

When caching is enabled, you can use the DELETE /caches actuator endpoint to clear all cache entries. It is enabled by default (management.endpoints.web.exposure.include=caches, ...).

IMPORTANT: in case you deploy the input api and the output api separately (using profiles), YOU NEED TO USE AN EXTERNAL CACHE (i.e. redis). This is because the output API is responsible for evicting old entries from the cache. If the input and the output do not run in the same JVM, the cache strategy simple is dangerous: the input api won't see if a token gets deleted, or if an object becomes disabled.

Current caching strategy:

  • one cache is used with the name metas,
  • metadata needed by the input api (unit, object owner, object state, object type, etc) get cached with a key objectId:token,
  • the cache is not updated when an object name or description changes (not really used by any downstream steps),
  • a single cache entry gets evicted when a token is deleted,
  • the entire cache is flushed when an object changes state (enabled/disabled), this is because we use compound keys, and have no way of using wildcard (e.g. objectId:*) in @CacheEvict. However, this operation is usually rare.

Async

To speed up the input endpoints, object statistics are updated asynchronously. Under the hood, a TheadPoolTaskExecutor is configured through the spring.task.execution.* properties (see application.properties in this repo for default values). You can of course override any of those in your properties file.

If you want to TURN OFF asynchronous processing, simply set the custom property async.enabled=false.

Logging

NOTE: in the default application.properties, I set the following, wich removes the spring banner, forces the use of ansi colors, so the output is highlighted even when using tee and always log resolved exceptions (feel free to change them if you need to):

spring.main.banner-mode=off
spring.output.ansi.enabled=ALWAYS
spring.mvc.log-resolved-exception=true

By default, Spring Boot uses logback. Here are some relevant links and resources:

Typically, in production you want to save everything to a rolling file, and maybe use syslog. You also want to be able to change the logging configuration without restarting the app. Here is an example on how to achieve that:

<?xml version="1.0" encoding="UTF-8"?>
<!-- FILE: logback-spring.xml -->
<!-- scan and scanPeriod define if and how often this configuration should be scanned for change at runtime -->
<configuration scan="true" scanPeriod="1 minutes">

    <!-- override default springboot properties -->
    <!-- To see which are available, scan throught the files at
         https://github.com/spring-projects/spring-boot/tree/master/spring-boot-project/spring-boot/src/main/resources/org/springframework/boot/logging/logback -->
    <property name="LOG_FILE" value="./logs/bbdata-api.log" />
    <property name="LOG_EXCEPTION_CONVERSION_WORD" value="%rEx{5}" />

    <!-- use Spring default values, making CONSOLE and FILE readily available -->
    <include resource="org/springframework/boot/logging/logback/base.xml" />

    <!-- Add syslog logging TODO: ensure you change the host and port ! -->
    <appender name="SYSLOG" class="ch.qos.logback.classic.net.SyslogAppender">
        <syslogHost>127.0.0.1</syslogHost>
        <facility>SYSLOG</facility>
        <port>514</port>
        <throwableExcluded>false</throwableExcluded>
        <!--<suffixPattern>bbdata %m thread:%t priority:%p category:%c exception:%exception</suffixPattern>-->
        <suffixPattern>bbdata: %p [%logger, %t] %m%n${LOG_EXCEPTION_CONVERSION_WORD}</suffixPattern>
    </appender>

    <!-- LOG everything at INFO level -->
    <root level="info">
        <appender-ref ref="CONSOLE" />
        <appender-ref ref="FILE" />
        <appender-ref ref="SYSLOG" />
    </root>

    <!-- LOG app at TRACE level -->
    <logger name="ch.derlin.bbdata" level="trace" additivity="false">
        <appender-ref ref="CONSOLE" />
        <appender-ref ref="FILE" />
        <appender-ref ref="SYSLOG" />
    </logger>

</configuration>

In order to load this configuration at runtime, use the option logging.config, for example:

# either using java
java -jar -Dlogging.config=logback-spring.xml
# or executing the jar directly
./bbdata-api-*.jar --logging.config=logback-spring.xml

rsyslog configuration

When using SYSLOG, you need a server with rsyslog running. To allow incoming udp requests, define the following in /etc/rsyslog.conf:

$ModLoad imudp
$UDPServerRun 514

To filter everything from bbdata create a file /etc/rsyslog.conf/10-bbdata.conf and add the following:

if $programname == 'bbdata' then /var/log/bbdata.log
& stop

The rule will match any log beginning with bbdata: (the default TAG defined in SYSLOG, see <suffixPattern>TAG: ...).

It is also possible to put the rule directly in rsyslog.conf at the beginning of the rules. If you put it at the end, everything will be logged in messages as well.

To deal with tabs, new lines and special characters, also define the following property in /etc/rsyslog.conf:

$EscapeControlCharactersOnReceive off

Permission system

‼️ tldr; IMPORTANT Ensure that the userGroup with ID 1 has a meaningful name in the database (e.g. "admin") and that you only add the platform managers to it !

The permission system in BBData uses mappings between userGroups and objectGroups.

A user can be part of one or more userGroups, either as regular user or as admin. Objects can be part of one or more objectGroups. ObjectGroups have a list of allowed userGroups (defining read permissions to objects that belong to it), and a single owner, which is the userGroup that created it. Only admin users from the owning userGroup can manage objects and permissions of an objectGroup. Regular users can access them in read mode. This is the same for objects: admins of the owning group has write access (e.g. editing metadata), regular users only read access (e.g. getting values).

There is one special case: the userGroup with userGroupId=1 is the ⚜️ SUPERADMIN ⚜️. This is the equivalent of SUDO: any admin of this group has read/write access to ANY resource, in read and write mode.

(see the documentation for more info)

Actuators and metrics

See Spring Boot Actuator: Production-ready Features.

Actuators are a way to monitor the API. They can leak sensitive information, so the management interface should run on another port as the API, which only administrators have access to.

By default, the actuators will run on the same port (easier for debug, as they will show in openapi), but this is definitely unsecure in production. Hence, ensure you select another port for the management interface by setting either:

# 8111 should be protected by firewall or not exposed to the outside
management.server.port=8111

or by disabling unsecure actuators, e.g.:

management.endpoints.web.exposure.include=info

Note that if you want to use one or more monitoring systems (see Monitoring), you will need to expose all metrics.

Customizing the /about (/info) endpoint

The actuator endpoint /info is mirrored in the public /about endpoint. What is actually displayed can be customized using properties.

By default, SpringBoot actuator will add the to the JSON response any property with the prefix info. For example, to configure the instance-name, define the following in your application.properties:

info.instance-name=My Super Instance

Another really cool feature is the possibility to reference other properties files. For example, let's say you want to show the current cache type in use, which is defined somewhere (in this or another properties file) using spring.cache.type=XX. To show this value in /info, simply add a property with the prefix dynamic.info (nested properties, with dots, are not allowed):

dynamic.info.cache-type=spring.cache-type

The JSON key cache-type will then appear in the /info endpoint. The pattern is thus:

dynamic.info.<JSON-KEY-NAME-WITHOUT-DOTS>=<ANOTHER PROPERTY>

To hide a dynamic property that is displayed, simply redefine it with an empty value, e.g. dynamic.info.cache-type=.

Task executor actuator

By default, and if sync.enabled=true, there is a custom /tasks actuator that returns statistics about the task executor (core size, active threads) and the tasks. To disable this endpoint, change exposed actuators (id=tasks).

Custom metrics

Custom metrics are:

  • Task Executor:

    • async.executor
    • async.executor.active
    • async.executor.completed
    • async.executor.idle
    • async.executor.pool.core
    • async.executor.pool.max
    • async.executor.pool.size
    • async.executor.queue.remaining
    • async.executor.queued
  • Input:

    • ch.derlin.input.rejected
    • ch.derlin.login.failed
    • ch.derlin.auth.failed

IMPORTANT: as long as there is no value, the metric won't appear / doesn't exist. This means /metrics/ch.derlin.login.failed will return 404 as long as no failed login occurred. Likewise, metrics/async_* won't appear if sync.enabled=false and/or no value has ever been submitted.

Changing exposed actuators

Spring Boot Actuator

By default, the exposed actuators are defined in application.properties, using the key management.endpoints.web.exposure.include=. Feel free to change this list as you see fit, or enable all using *.

To turn off all endpoints, simply add:

management.server.port=
management.endpoints.web.exposure.include=none

Monitoring

There are two (complementary?) supported monitoring systems:

  • Spring Boot Admin, which lets you interact with the actuators from a UI (clear the cache, change logging levels, etc.);
  • Prometheus + Grafana, which lets you graph and monitor different metrics.

Spring Boot Admin

Spring Boot Admin needs a server running, as well as a client embedded into the API.

For a basic server implementation, see https://github.com/big-building-data/spring-boot-admin (nightly release).

The client is already included in the bbdata-api jar. What you need to do:

  1. run a Spring Boot Admin server,
  2. set the properties below in your application.properties
## Spring Boot Admin

# expose every actuator available, but ensure it runs on another (secured) port !
management.server.port=8111
management.endpoints.web.exposure.include=*

# enable spring boot admin client, and provide the server's URL
spring.boot.admin.client.enabled=true
spring.boot.admin.client.url=<URL OF THE ADMIN SERVER>
spring.boot.admin.client.instance.name=BBData test Instance

Prometheus + Grafana

The BBData API ships with io.micrometer:micrometer-registry-prometheus. If the property management.endpoint.prometheus.enabled=true and prometheus is exposed (see management.endpoints.web.exposure.include), you can see every available metric at http://localhost:8111/prometheus.

See the other/monitoring's README for more info on how to setup Prometheus and Graffana.