My experience with Strigo is that all processes are killed when Strigo is paused in any way (either automatically or manual pause). As a result all data ingest etc is stopped.
This happens despite that they are started as background processes (&).
The lab documentation on the other hand, states that the processes shall be kept running.
The problem is very time consuming as trying to achieve ch 6 labs relying on earlier processes.
The question is then how to maintain processes in Strigo?
Maintaining the beats has proven to be difficult for this course in particular, and we haven't really found an effective solution so far. So, it's a manual process that each student will have to start up the beats on various servers for various services each time.
I would argue that the documentation should be updated in this case to present the pre-processes to be started. Not that each student will have to be Sherlock Holmes to track what has to be executed. Potentially the beats are quite straight forward but then there are various machines etc. The APM is more complex as well.
It is then almost impossible to execute the ML labs as sufficient data is required for ML which cannot be maintained due to these limitations. It thus seems that ch 6 is out of context.
As to the ML labs, ML anomaly detection requires, of course, anonmalies to detect. Even in a live delivery, with data being collected over a couple days, we sometimes get no anomalies, because the items we're collecting data about are relatively stable and generally don't spike, for instance, CPU usage, or ping times, or hosts dropping for an hour. If you do the lab and understand the processs for creating an ML job, then you've gotten about 90% of the value of the lab.
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.