Module 4 — Exploring Managed data: hosts, services, Dynatrace PurePath and problems

Explore Managed data step by step: hosts and services, service flow, Dynatrace PurePath distributed traces and reading a Davis problem, with exercises.
Module 4 — Explorer : données Managed, hôtes, services, PurePath Dynatrace et problèmes

Module 4 of the Dynatrace Managed course: reading hosts, services, Dynatrace PurePath traces and problems in the environment interface.

Course: Dynatrace Managed · Module 4/8 · Duration: ~45 min · Level: Beginner to intermediate

Objectives:

  • Navigate from a host to its processes and services
  • Read a service’s key indicators and find its slow requests
  • Analyse a PurePath distributed trace
  • Read a problem view and trace it back to the root cause

Prerequisites: module 3; at least one host with an instrumented application receiving traffic.

Your agents are sending data: now you need to know how to read it. This module teaches you how to explore Dynatrace Managed data in an environment’s interface: from hosts to services, from services to Dynatrace PurePath traces, then the view of a problem opened by Davis. This is the skill you will use most often day to day.

Exploring Managed data, starting from the host

A host page brings together its system metrics (CPU, memory, disk, network), recent events and the list of its processes. Get into the habit of reading it from top to bottom:

  1. Availability and events: restart, full disk, stopped process.
  2. Resource consumption: saturation often explains an application slowdown.
  3. Processes: which processes consume the most, which ones are deeply instrumented.

From a process, you reach its process group and the services it hosts. Here you are following the vertical Smartscape axis from module 1, but from the bottom up.

How do you analyse a service?

A service is a technical entry point detected automatically: a web application, a REST API, a messaging service, a database as seen by its callers. Its page shows three main indicators:

Indicator What to look at
Response time The median and the high percentiles: slowness often hides in the “tail” of the distribution
Failure rate The codes and exceptions that make requests fail
Throughput The number of requests per minute, to tell a load increase from a degradation

The service’s list of requests ranks entry points (URLs, methods) by response time or volume. To find out where the time goes, two analyses are invaluable:

  • Response time analysis breaks the duration down between the service’s own code, calls to other services and databases.
  • Service flow shows the chain of downstream calls from this service, with each step’s share of time and failures. Conversely, the backtrace shows who calls this service.

Reading a Dynatrace PurePath distributed trace

PurePath is the name of Dynatrace distributed traces. A trace follows a single request through all the services and processes it passes through. It is shown as a tree:

GET /api/orders/42                  web-frontend       820 ms
 +- GET /orders/42                  orders-service     790 ms
     +- SELECT ... FROM order       orders-db           35 ms
     +- POST /stock/check           stock-service      710 ms
         +- external HTTP call      supplier           690 ms

Here, the reading is immediate: most of the 820 ms comes from an external call made by the stock service. For each node in the tree, PurePath provides, depending on the technology, the request attributes, parameters, exceptions and, at the finest level, the breakdown of time by method (CPU time, waiting, locks, I/O).

Recommended reading method:

  1. Filter traces on the slow or failing request, and on the incident time frame.
  2. Open a few representative traces, not just the worst one.
  3. Find the node where the time is concentrated, then drill into its details.
  4. Match any error with the logs of the process concerned.

Traces are stored on the cluster nodes, with a limited retention period: their retention can be configured and sized, as explained in Data retention and storage management in Managed.

Understanding a problem view

When Davis opens a problem, the problem page is the starting point of the investigation. It contains:

  • The title and category: availability, error, slowdown, resource, custom event.
  • The impact: applications, services and, when RUM is enabled, affected users.
  • The probable root cause: the entity and event that Davis places at the origin of the chain.
  • The evolution over time (visual resolution path): the order in which entities were affected.
  • Comments and integrations: team notes, notifications sent.

Always read the root cause as a strong hypothesis, then confirm it by opening the traces or metrics of the entity identified. Module 5 explains how to tune detection and who to send these problems to.

Hands-on exercise

Do this on your test environment, with the application instrumented in module 3.

Step 1 — Walk the vertical stack

Open the host page, then an instrumented process, then the service it hosts. Expected result: you can name your application’s host, process group and service.

Step 2 — Create a slow request

Add a page to your test application that waits two seconds (or call a slow URL), then generate traffic mixing fast and slow requests. Expected result: the slow request appears at the top of the request list sorted by response time.

Step 3 — Analyse a trace

Open a PurePath trace of the slow request and identify the node where the time is concentrated. Expected result: a one-sentence written explanation (“2 s of waiting in such-and-such method”).

Step 4 — Trigger and read a problem

Stop the monitored process for a few minutes or make most requests fail, then wait for a problem to open. Expected result: a problem with its impact and root cause; note the delay between the incident and the problem opening, then restore the service.

Quiz

  1. What does the service flow show? a) The cluster hosts b) The chain of downstream calls from a service c) The service logs d) The users signed in
  2. Why look at high percentiles and not just the median? a) The median does not exist in Dynatrace b) The slowness experienced by some users hides in the tail of the distribution c) Percentiles cost less d) Davis only reads percentiles
  3. What is a PurePath trace? a) An aggregated metric b) The complete path of a single request through the services c) A log file d) A licence report
  4. How should you treat the root cause shown in a problem? a) As an absolute certainty b) As a strong hypothesis to be confirmed with traces and metrics c) As a mere suggestion of no value d) By ignoring it if several services are affected

Answers

  1. b) The service flow breaks down downstream calls, with the time and failures of each step.
  2. b) A reasonable median can hide very slow requests for a minority of users.
  3. b) PurePath follows a request end to end, down to method-level detail.
  4. b) Davis relies on the topology, but checking against the data is still the right reflex.

Further reading

← Module 3: Deploying OneAgent and ActiveGate · Course overview · Module 5: Davis, problems and alerting →

Frequently asked questions

How long are PurePath traces kept?

It depends on the configuration and the cluster’s disk space. In Managed, the administrator sets the retention; check the default values and limits in the documentation for your version.

Why does my database appear as a service with no agent on it?

Dynatrace detects databases from the calls made by instrumented processes. You therefore see the queries and their timings as the applications experience them, even with no agent on the database server.

Is every request traced?

Not necessarily one by one: at high volume, intelligent sampling applies to limit the load while keeping reliable statistics. The details depend on the version and the volume.

You now know how to go from a symptom to its cause: from host to service, from service to trace, from problem to root cause. Module 5 shows you how to tune Davis and get the right problems to the right teams.

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
2 comments
Leave a Reply

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

Previous Post
Module 3 — OneAgent, ActiveGate Dynatrace : les déployer dans Managed

Module 3 — Deploy OneAgent and ActiveGate in Dynatrace Managed

Next Post
Module 5 — Davis : alerting en Managed, anomalies, problèmes et profils d'alerte

Module 5 — Dynatrace Davis alerting in Managed: anomalies, problems and alerting profiles

Related Posts