Module 1 — Observability concepts: APM, OneAgent, Smartscape and Davis

Observability concepts for getting started with Dynatrace Managed: APM, OneAgent, Smartscape topology and Davis root cause, with an exercise and a quiz.
Module 1 — Concepts : observabilité, APM, OneAgent, Smartscape et Davis

Module 1 of the Dynatrace Managed course covers the observability concepts (APM, OneAgent, Smartscape and Davis) that every later module builds on.

Course: Dynatrace Managed · Module 1/8 · Duration: ~35 min · Level: Beginner

Objectives:

  • Distinguish classic monitoring, APM and observability
  • Explain the role of OneAgent and what it collects
  • Read the Smartscape topology and its two axes
  • Understand how Davis goes from anomalies to a root cause

Prerequisites: no knowledge of Dynatrace; a basic understanding of web application architecture.

Before installing anything, you need a shared vocabulary. This first module of the course sets out the observability concepts as Dynatrace Managed implements them: APM, the single OneAgent, the Smartscape dependency map and the Davis analysis engine. These four ideas come back in every following module, so take the time to master them.

Observability concepts: monitoring, APM or observability?

Classic monitoring answers questions known in advance: is the disk full? Does the service answer the ping? You define thresholds and receive alerts. It works well as long as failures look like the ones you imagined.

APM (Application Performance Monitoring) goes down into the application: response time per request, error rate, database calls, slow methods. You no longer look only at the machine, but at the work it does for the user.

Observability goes further: the ability to understand behaviour nobody anticipated, by querying rich, interconnected data. Three pillars are often cited, to which Dynatrace adds context:

Signal Question it answers Example
Metrics How much? How often? CPU of a host, median response time of a service
Traces Which path did this request take? An HTTP call that goes through a front end, two microservices and a database
Logs What did the component write at that moment? A Java exception with its stack trace
Topology What depends on what? This service runs in this process, on this host, and calls this database

The strength of a platform such as Dynatrace is that it links these signals automatically: a slow trace is attached to the service, process and host that executed it.

OneAgent: a single agent to collect everything

OneAgent is an agent installed once per host (physical server, virtual machine, Kubernetes node). Once installed, it:

  • collects operating system metrics (CPU, memory, disk, network);
  • automatically discovers the processes running on the host and groups them;
  • injects itself into processes of supported technologies (Java, .NET, Node.js, PHP, web servers…) to trace requests: this is deep monitoring;
  • can collect logs and inject RUM (real user monitoring) into the web pages served.

The key point: no code changes are needed for supported technologies. On the other hand, a process that was already running before the agent was installed must be restarted to be instrumented. You will check this in module 3.

In Managed, OneAgent sends its data to the cluster installed on your premises, either directly or through an ActiveGate acting as a relay. This flow is described in detail in the article Understanding Dynatrace Managed architecture.

Smartscape: the live dependency map

From what the agents discover, Dynatrace builds a topology model called Smartscape. It is organised in stacked layers:

Applications   (what the user sees: pages, actions)
     |
Services       (technical entry points: REST APIs, servlets, SQL queries)
     |
Processes      (JVM, IIS, nginx, Node.js… grouped into process groups)
     |
Hosts          (servers, VMs, nodes)
     |
Datacenters    (sites, zones, cloud regions)

Smartscape is read along two axes:

  • Vertical: “this service runs in this process, on this host”. This is the hosting relationship.
  • Horizontal: “this service calls that service”, “this process talks to that one”. This is the call relationship.

This map is not drawn by hand: it updates continuously as deployments happen. It is what allows Davis to reason.

Entities: the basic vocabulary

Each element in Smartscape is an entity with a unique identifier: host, process group, process instance, service, application. The tags, management zones and APIs you will see in modules 6 and 8 apply to these entities. Remember the difference between a process group (for example “all the JVMs of the orders application”) and a process instance (one specific JVM on one specific host).

Davis: from anomaly to root cause

Davis is the Dynatrace artificial intelligence engine. Its work happens in three stages:

  1. Learn normal behaviour: baselines are calculated automatically for response times, error rates and traffic of services and applications.
  2. Detect anomalies: a significant deviation from the baseline, or a fixed threshold being exceeded for resources (full disk, stopped process), produces an event.
  3. Correlate: instead of sending ten alerts, Davis walks up Smartscape to link the events together and opens a single problem, with a probable root cause and the impact on users.

Example: a database fills up its disk, three services that call it slow down, and the web application’s error rate climbs. Classic monitoring sends five alerts; Davis presents a single problem, identifies the database as the root cause and quantifies the impact. Module 5 is dedicated to configuring this detection and routing alerts.

Hands-on exercise

This exercise can be done on paper or on a Managed demo environment if you have read access to one.

Step 1 — Draw your own Smartscape

Pick an application you know. List its hosts, its processes (web server, application server, database), its services and the calls between them. Place them on the layers of the diagram above. Expected result: a diagram with four or five layers, vertical arrows (hosting) and horizontal arrows (calls).

Step 2 — Classify the signals

For three recent incidents you have dealt with, note which signal (metric, trace, log, topology) would have found the cause fastest. Expected result: at least one incident where the topology or the trace would have been decisive.

Step 3 — Imagine the Davis correlation

Take one of those incidents and write down the alerts you received. Group them as Davis would, and identify the root cause by walking up the dependencies. Expected result: several alerts reduced to one problem and one root entity.

Step 4 — Explore a demo (optional)

On a demo environment, open the Smartscape view and click a service: find its process, its host and the services it calls. Expected result: the two reading axes found in the tool.

Quiz

  1. Which concept answers the question “which path did this request take?”? a) Metric b) Trace c) Log d) Baseline
  2. How many OneAgents do you install per host? a) One per technology b) One per process c) Just one d) None, everything goes through ActiveGate
  3. In Smartscape, the relationship “this service runs in this process” belongs to the: a) horizontal axis b) vertical axis c) time axis d) application axis
  4. What does Davis do when several related anomalies occur? a) It sends one alert per anomaly b) It ignores them until manually validated c) It correlates them into a problem with a probable root cause d) It restarts the processes involved

Answers

  1. b) A trace follows a request end to end across services.
  2. c) A single OneAgent per host covers the system and the supported processes.
  3. b) Hosting (service, process, host) is read vertically; calls are read horizontally.
  4. c) Davis relies on the topology to group events into a single problem.

Further reading

← Course overview · Module 2: Managed cluster architecture and installation →

Frequently asked questions

Are observability and APM the same thing?

No. APM measures application performance; observability includes APM and adds the correlation of metrics, traces, logs and topology to understand unexpected behaviour.

Do I need to change the code for OneAgent to trace requests?

Not for supported technologies: the agent injects itself when the process starts. You do, however, need to restart processes launched before the installation. Check the compatibility matrix for your version.

Does Davis replace alert thresholds?

Largely, for services and applications, thanks to automatic baselines. Fixed thresholds remain useful for some resources and for specific commitments, such as a contractual response time.

You now have the course vocabulary: metrics, traces, logs, entities, Smartscape and Davis problems. The next module shows you where this data is stored and how to install the cluster that receives it.

span.observer is an independent community blog, not affiliated with Dynatrace LLC. This course does not replace the official Dynatrace University training and certifications. Check features in the official documentation for your version.

Total
0
Shares
Leave a Reply

Your email address will not be published. Required fields are marked *

Previous Post
Formation Dynatrace Managed : le parcours complet en 8 modules

Dynatrace Managed training: a free Dynatrace course in 8 modules

Next Post
Formation Managed module 2 : nœuds, stockages et installation d'un premier cluster de test

Module 2 — Dynatrace Managed cluster: nodes, storage and a first test cluster

Related Posts