Skip to content

Latest commit

 

History

History
220 lines (154 loc) · 10.7 KB

installation.md

File metadata and controls

220 lines (154 loc) · 10.7 KB
Raw

Installation

In order to set up your very own Pinpoint instance, you need to run these components:

  • HBase (for storage)
  • Pinpoint Collector (deployed on a web container)
  • Pinpoint Web (deployed on a web container)
  • Pinpoint Agent (attached to a java application for profiling)

To try out a simple quickstart project, please refer to the quick-start guide.

Quick Overview

  1. HBase (details)
    1. Set up HBase cluster - Apache HBase
    2. Create HBase Schemas - feed /scripts/hbase-create.hbase to hbase shell.
  2. Build Pinpoint (Only required if you're building from source)
    1. Clone Pinpoint - git clone $PINPOINT_GIT_REPOSITORY
    2. Set JAVA_6_HOME environment variable to JDK 6 home directory.
    3. Set JAVA_7_HOME environment variable to JDK 7+ home directory.
    4. Run mvn install -Dmaven.test.skip=true in Pinpoint home directory.
  3. Pinpoint Collector (details)
    1. Deploy pinpoint-collector-$VERSION.war to a web container.
    2. Configure pinpoint-collector.properties, hbase.properties.
    3. Start container.
  4. Pinpoint Web (details)
    1. Deploy pinpoint-web-$VERSION.war to a web container as a ROOT application.
    2. Configure pinpoint-web.properties, hbase.properties.
    3. Start container.
  5. Pinpoint Agent (details)
    1. Extract/move pinpoint-agent/ to a convenient location ($AGENT_PATH).
    2. Set -javaagent:$AGENT_PATH/pinpoint-bootstrap-$VERSION.jar JVM argument to attach the agent to a java application.
    3. Set -Dpinpoint.agentId and -Dpinpoint.applicationName command-line arguments.
    4. Launch java application with the options above.

HBase

Pinpoint uses HBase as its storage backend for the Collector and the Web.

To set up your own cluster, take a look at the HBase website for instructions. The HBase compatibility table is given below:

Pinpoint Version HBase 0.94.x HBase 0.98.x HBase 1.0.x HBase 1.1.x
1.0.x yes no no no
1.1.x no not tested yes not tested
1.5.x no not tested yes not tested

Once you have HBase up and running, make sure the Collector and the Web are configured properly and are able to connect to HBase.

Creating Schemas

There are 2 scripts available to create tables for Pinpoint: hbase-create.hbase, and hbase-create-snappy.hbase. Use hbase-create-snappy.hbase for snappy compression (requires snappy), otherwise use hbase-create.hbase instead.

To run these scripts, feed them into the HBase shell like below:

$HBASE_HOME/bin/hbase shell hbase-create.hbase

See here for a complete list of scripts.

Building Pinpoint

There are two options:

  1. Download the build results from our latest release and skip the building. Recommended.

  2. Build Pinpoint manually from the Git clone.

    In order to do so, the following requirements must be met:

    • JDK 6 installed
    • JDK 7+ installed
    • Maven 3.2.x+ installed
    • JAVA_6_HOME environment variable set to JDK 6 home directory.
    • JAVA_7_HOME environment variable set to JDK 7+ home directory.

    JDK 7+ and JAVA_7_HOME environment variable are required to build profiler-optional. For more information about the optional package, please take a look here.

    Additionally, the required Java version to run each Pinpoint component is given below:

    Pinpoint Version Agent Collector Web
    1.0.x 6+ 6+ 6+
    1.1.x 6+ 7+ 7+
    1.5.x 6+ 7+ 7+

    Once the above requirements are met, simply run the command below :

    mvn install -Dmaven.test.skip=true

    The guide will refer to the full path of the pinpoint home directory as $PINPOINT_PATH.

Regardless of your method, you should end up with the files and directories mentioned in the following sections.

Pinpoint Collector

You should have the following war file that can be deployed to a web container.

pinpoint-collector-$VERSION.war

The path to this file should look like $PINPOINT_PATH/collector/target/pinpoint-collector-$VERSION.war if you built it manually.

Installation

Since Pinpoint Collector is packaged as a deployable war file, you may deploy them to a web container as you would any other web applications.

Configuration

There are 2 configuration files available for Pinpoint Collector: pinpoint-collector.properties, and hbase.properties.

  • pinpoint-collector.properties - contains configurations for the collector. Check the following values with the agent's configuration options :
    • collector.tcpListenPort (agent's profiler.collector.tcp.port - default: 9994)
    • collector.udpStatListenPort (agent's profiler.collector.stat.port - default: 9995)
    • collector.udpSpanListenPort (agent's profiler.collector.span.port - default: 9996)
  • hbase.properties - contains configurations to connect to HBase.
    • hbase.client.host (default: localhost)
    • hbase.client.port (default: 2181)

These files are located under WEB-INF/classes/ inside the war file.

You may take a look at the default configuration files here: pinpoint-collector.properties, hbase.properties

Pinpoint Web

You should have the following war file that can be deployed to a web container.

pinpoint-web-$VERSION.war

The path to this file should look like $PINPOINT_PATH/web/target/pinpoint-web-$VERSION.war if you built it manually.

Installation

Since Pinpoint Web is packaged as a deployable war file, you may deploy them to a web container as you would any other web applications. The web module must also be deployed as a ROOT application.

Configuration

Similar to the collector, Pinpoint Web has configuration files related to installation: pinpoint-web.properties, and hbase.properties.

Make sure you check the following configuration options :

  • hbase.properties - contains configurations to connect to HBase.
    • hbase.client.host (default: localhost)
    • hbase.client.port (default: 2181)

These files are located under WEB-INF/classes/ inside the war file.

You may take a look at the default configuration files here: pinpoint-web.properties, hbase.properties

Pinpoint Agent

If downloaded, unzip the Pinpint Agent file. You should have a pinpoint-agent directory with the layout below :

pinpoint-agent
|-- boot
|   |-- pinpoint-bootstrap-core-$VERSION.jar
|-- lib
|   |-- pinpoint-profiler-$VERSION.jar
|   |-- pinpoint-profiler-optional-$VERSION.jar
|   |-- pinpoint-rpc-$VERSION.jar
|   |-- pinpoint-thrift-$VERSION.jar
|   |-- ...
|-- pinpoint-bootstrap-$VERSION.jar
|-- pinpoint.config

The path to this directory should look like $PINPOINT_PATH/agent/target/pinpoint-agent if you built it manually.

You may move/extract the contents of pinpoint-agent directory to any location of your choice. The guide will refer to the full path of this directory as $AGENT_PATH.

Installation

Pinpoint Agent runs as a java agent attached to an application to be profiled (such as Tomcat).

To wire up the agent, pass $AGENT_PATH/pinpoint-bootstrap-$VERSION.jar to the -javaagent JVM argument when running the application:

  • -javaagent:$AGENT_PATH/pinpoint-bootstrap-$VERSION.jar

Additionally, Pinpoint Agent requires 2 command-line arguments in order to identify itself in the distributed system:

  • -Dpinpoint.agentId - uniquely identifies the application instance in which the agent is running on
  • -Dpinpoint.applicationName - groups a number of identical application instances as a single service

Note that pinpoint.agentId must be globally unique to identify an application instance, and all applications that share the same pinpoint.applicationName are treated as multiple instances of a single service.

Tomcat Example

Add -javaagent, -Dpinpoint.agentId, -Dpinpoint.applicationName to CATALINA_OPTS in the Tomcat startup script (catalina.sh).

CATALINA_OPTS="$CATALINA_OPTS -javaagent:$AGENT_PATH/pinpoint-bootstrap-$VERSION.jar"
CATALINA_OPTS="$CATALINA_OPTS -Dpinpoint.agentId=$AGENT_ID"
CATALINA_OPTS="$CATALINA_OPTS -Dpinpoint.applicationName=$APPLICATION_NAME"

Start up Tomcat to start profiling your web application.

Configuration

There are various configuration options for Pinpoint Agent available in $AGENT_PATH/pinpoint.config.

Most of these options are self explanatory, but the most important configuration options you must check are collector ip address, and the TCP/UDP ports. These values are required for the agent to establish connection to the Collector and function correctly.

Set these values appropriately in pinpoint.config:

  • profiler.collector.ip (default: 127.0.0.1)
  • profiler.collector.tcp.port (collector's collector.tcpListenPort - default: 9994)
  • profiler.collector.stat.port (collector's collector.udpStatListenPort - default: 9995)
  • profiler.collector.span.port (collector's collector.udpSpanListenPort - default: 9996)

You may take a look at the default pinpoint.config file here along with all the available configuration options.

Miscellaneous

Routing web requests to agents

Starting from 1.5.0, Pinpoint can send requests from the web to agents directly via the collector (and vice-versa). To make this possible, we use Zookeeper to co-ordinate the communication channels established between agents and collectors, and those between collectors and web instances. With this addition, real-time communication (for things like active thread count monitoring) is now possible.

We typially use the Zookeeper instance provided by the HBase backend so no additional Zookeeper configuration is required. Related configuration options are shown below.

  • Collector - pinpoint-collector.properties
    • cluster.enable
    • cluster.zookeeper.address
    • cluster.zookeeper.sessiontimeout
    • cluster.listen.ip
    • cluster.listen.port
  • Web - pinpoint-web.properties
    • cluster.enable
    • cluster.web.tcp.port
    • cluster.zookeeper.address
    • cluster.zookeeper.sessiontimeout
    • cluster.zookeeper.retry.interval
    • cluster.connect.address