We are developing a kind of progressive web app, and 90% of the code is run in the browser.
During last year we used RUM APM with custom transactions to trace all the domain code and get performance insights. Right now, we want to configure some alerts based on those metrics to detect when the performance degrades, but we haven't been able to use the APM alerts since the metrics are very different depending on the user computer specs, and the amount of data the user requires.
We tested all the rules based on APM, and also tried Elasticsearch based rules, but the query is too complex since requires comparing data with previous dates and we don't know how to do it.
Are we missing something? Are those kind of alerts possible in our context?
Important to highlight that RUM data isn't currently integrated with Alerting and we don't have a defined timeline for when this integration is going to be done.
Just to clarify my question, we are using @elastic/apm-rum integration with custom transactions where we append relevant information not related with RUM data.
We want to detect performance regressions on those transactions based on the custom data we appended, not with RUM data Elastic agent provides.
We tested the Latency anomaly and Latency threshold rule types, but since the performance of those transactions depend on the user and the amount of data of that user, the transaction performance is not comparable. When using Log threshold we have been able to solve that by grouping by our metadata, but with this APM rules we can't.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.