April 30, 2020

Monitoring AWS Analytics

Monitoring AWS Analytics

Amazon Web Services (AWS) products are countless, and at LogicMonitor, we are working tirelessly to bring monitoring support to as many of them as possible. With so many products and tools already on your plate, we want to make sure that monitoring is not a hassle, but rather a trusted companion. Here, we will focus on the analytics section of AWS and provide some tips on how to utilize the data collected from AWS Athena and Glue.

AWS Athena Success Dashboard

AWS Athena

Amazon Athena is an interactive query service that analyzes data stored in Amazon Simple Storage Service (S3) using standard SQL. By monitoring Athena, we give users insight into failed queries and long-running queries. This allows them to catch issues sooner and resolve them faster, thus maximizing uptime and performance.

The data is collected into two query types: Data Definition Language (DDL) and Data Manipulation Language (DML). Each extracted metric is identified by the state of the query: canceled, failed, and succeeded. With that indicator, you can focus your attention on the metrics that are relevant to your task. You can also create a dashboard specifically for failed queries, which is useful for troubleshooting, and another dashboard for successful queries, which will help you come up with acceptable baselines for more accurate thresholds. While troubleshooting, you want to track the execution time. By bringing all the time metrics such as query queue, query planning, engine execution, and service processing into one place, you can quickly identify bottlenecks or the place where things break down. Once you feel like you are in a steady-state and things are running smoothly, make sure to go back and set suitable thresholds for the time metrics. With this, along with our dynamic thresholds feature, you can rest assured that you will only get notified when it matters.

AWS Glue

AWS Glue is a fully managed extract, transform, and load (ETL) service that helps customers prepare and load their data for analytics. The service automatically detects an enterprise’s structured or unstructured data when it is stored within other AWS services, such as data lakes in Amazon Simple Storage Service (S3), data warehouses in Amazon Redshift, and other databases that are part of the Amazon Relational Database Service. All of these can be monitored by LogicMonitor. The only objects from AWS Glue with measurable data are the Jobs, so this DataSource discovers the jobs automatically.

To accurately monitor Jobs, we need to keep track of executors, stages, and tasks.

The number of actively running job executors, “active executors”, can be helpful to identify any delays due to straggler scenarios. Keep an eye on active executors and the max executors metrics to know when it is time to provision more data processing units (DPUs). Make sure to create an alert and warnings for active executors metric as it approaches the max executors value. 

As your job goes through different stages, it is helpful to have the completed stages metric on a dashboard to monitor the progress. If you notice that some stages are taking too long, you can correlate the CPU load, memory, and elapsed time metrics to triage the issue and potentially add more resources. Or you can repartition the load to lower the elapsed time. If there is any parallelism within your stages, it is essential to track the completed task and failed tasks. Depending on the workflow of the job, you need to set an appropriate threshold for failed tasks. These failed tasks can help determine abnormalities within the data, cluster, or script.

We recently highlighted our commitment to our AWS partnership, and our developers keep cranking out new AWS monitoring integrations. All of these new integrations are included with our LM Cloud offering. LogicMonitor will automatically start collecting these metrics if you are an LM Cloud customer and already utilizing any of these Amazon services. 
If you have any questions regarding these new monitoring integrations, please reach out or contact your CSM for more information. Or you can try LogicMonitor for free.

Juan Carlos Reyes

Juan Carlos Reyes

Juan Carlos is a loud and proud member of the Fightin’ Texas Aggie Class of 2014. He has 4+ years of experience as a performance engineer working with APM tools such as Dynatrace and New Relic. In his free time, he enjoys going on long hikes and exploring national parks with his wife. If you know any fun and exciting board games, make sure to drop him a note since he loves trying new games. Thanks and Gig’em!