General

Q1: How do I know when new features are added to APM?

Answer: All major new features are announced in the APM Release Notes


Q2: Is it possible to switch an APM Domain from Free tier to Paid?

Answer: No. You need to create a new paid domain and reconfigure agents/tracer/browser agent settings.


Q3: How can I limit the ingestion, so it will not go over a set number of billing units?

Answer: Limiting ingestion can be done in multiple ways:

  • Exclude unwanted or not useful spans. e.g. use the servlet probe configuration to exclude high volume - low value service request.
  • Use sampling to reduce the number of traces being reported. Note that the decision to sample a trace is done in the first component of a trace flow.
  • Reduce the abridge limit (default value is 100)

Q4: How can I be alerted on potential increase in service usage cost due to changes in application usage?

Answer: Yes, the metric SpanIngestions in the oci_apm namespace reports the number of span ingested to a domain in 15 minutes intervals. You can set an alert for this metric. For billing proposes the span count is hourly: either aggregating to 1h units ( SpanIngestions[1h].sum() ) or using ¼ of the desired threshold value. Note that 100,000 spans equal one billing unit.


Q5: Is APM reporting of username compliant with GDPR?

Answer: By default, no user identifiable information is captured. When username reporting functionality is enabled, special care needs to be taken to align usage with GDPR standards.


Monitoring

Q1: How are multiple browser tabs for a web application handled in monitoring?

Answer: By default, multiple tabs will be treated as a single session with the same session-id.


Q2: OCI Application Performance Monitoring provides only a Java agent for download and deployment. Does that mean that APM cannot monitor .NET or any other environments?

Answer: Not at all. APM analyzes data from a variety of sources, not just from Agents deployed in an environment. Among others, APM supports OpenTracing. Specifically, APM supports the JSON-encoded Zipkin v2 format. Using Jaeger, Zipkin or other tracers, APM can monitor most common languages.


Q3: The “Complete" and “Success” status of a trace seem very similar. What is the difference?

Answer: The “Complete” status indicates that the trace root span is complete. A “Success” status indicates that the root span is complete and that it has an http-status dimension, the value of which represents a successful request (0 < httpStatusCode < 400).


Q4: What's the difference between APM Java agent and APM Java Tracer?

Answer: The APM Java agent does byte code instrumentation (BCI) when added to an application server. It uses the open tracing format to send data to the APM end point. If you don’t have a way to generate tracing in the application server, you can use APM agent and its BCI to get traces. If your application server supports creating traces (for example Helidon or Spring Boot) you can use the APM Java tracer to send traces to APM. For more information, see Configure Application Performance Monitoring Data Sources.


Time Reporting

Q1: What is the difference between end time and report period?

Answer: The best way to think of report period is just to assume it's the end time for the span. This provides a consistent view of all received metrics for reporting and alerting. The main benefit of report period over an end time concept, is that the end time would be based on agent timestamp (start time + duration), and clock skew would make the resulting reporting less useful.


Q2: What is the impact of clock skew on reporting?

Answer: Clock skew refers to the fact that the clock at different agents is not synchronized. The result of clock skew is that the start time of different spans and the related duration does not match up when it's diagnosed in detail. This impacts metrics like trace duration (which is based on minimum start time of any span in the trace, vs maximum start-time + duration of any span in the trace).


Q3: What time can I use?

Answer: Common concepts can be used to identify which of the two timings to use:

  • When diagnosing details of a trace, it is often easiest to follow the tree by looking at the span start times, and reason from there, what triggered what
  • When trying to identify a trace that relates to data seen in a logfile captured at the same system the agent is running, using the span start time is the best aproach, since these are stemming from the same clock as the logfile.
  • When aggregating data to see load and performance metrics over time, it is best to use the reportPeriod, since this metric is most stable, especially when the agent is capturing long running processes.
  • When reporting a dashboard that reflects current state, it is best to use reportPeriod, since the current state of affairs is more complete.
  • When creating alerts, it is important to use the reportPeriod, to ensure client clock skew (especially from browsers) is not hiding any impact on issues.

Errors and Codes

Q1: Why are errors of type 'Script Error' reported without further details?

Answer: In some cases the CORS implementation in the browser is preventing access to details of javascript that caused a particular failure. It is suggested to move the apmrum javascript to the monitored application domain to improve visibility. In some cases adding the 'Timing-Allowed' header can be used to increase visibility in reporting.


Q2: What does http status code 0 indicate?

Answer: http status 0 indicates that there was no response from the server to derive the status code from, indicating either a timeout or restrictions in the browser that prevent access to this data (e.g. due to CORS).