How to use the Observation for Adobe Commerce nerdlet

Examine the % of Storage Free and MySQL % free storage by node frames.

• Follow the links in the frame’s header if you see low storage.

Examine the % of free system memory and Swap memory free in bytes frames.

• If these display very low memory states, they can be contributors to problems.

Examine the Alerts during the timeframe frame.

• Adobe Commerce on cloud infrastructure provides Managed alerts. You can click on the link in the header to see Support Knowledge Base articles that will help determine actions on your part for specific alerts.

Examine the CPU % by host frame: If it is exhibiting high CPU utilization, check the Support Knowledge Base article in the header for the frame. Also, check to ensure that database imports/exports or backups are not being done during peak traffic periods.

Check the Web Traffic volume compared to one week ago frame: If the traffic is much higher than the prior week during the same period, can it be explained (sale campaign or new products that have been marketed, for example)?

• If an increase in traffic can’t be explained, look at the Average Response time (milliseconds) for the production environment. Is the higher traffic contributing to a response time different than what is normal? Expand the timeframe to see if it is an anomaly.
• Is the increase in traffic impacting web transactions? Check the Response Code frame for errors. If the site is down, you can click on the Site Down? link in the frame header. The frame will identify any errors that are occurring and their frequency.
• Did someone deploy changes to your website? The Deployment Log Entries frame will indicate whether any deployments have been done during the issue timeframe. If the problem is immediately following deployment, it might be that deployment activities are adding additional load to the site (caches cleared, services restarted, etc.).
• Did an upsize or downsize occur? If your site was upsized temporarily, it might have been returned to its original cluster size. If a request was made to increase the site capacity, an upsize might occur. Check the Upsize/Downsize – vCPU view over the timeline frame. This frame will sometimes detect an outage on one particular node. If you see the size decrease, that may indicate an issue with one or more nodes.

The IP Frequency tab identifies request frequency from IP addresses that are made against the origin servers (meaning that the request could not be served from Fastly as 74 it wasn’t cached).

• For any Fastly related issues, check the Fastly Cache frame, and select the Error facet to see the percentage of requests that are errors. They may indicate a backend issue if they coincide with non-web load.
• If the load does not appear to be due to web traffic, there may be errors or a build-up of non-web requests, such as slow queries or crons.

Check the Database Errors frame for errors that may coincide with the issue/problem timeline.

Check the Database mysql-slow.log frame to identify SQL statements that are occurring. INSERT, UPDATE, and DELETE commands may take a while if the query is not optimized. Even SELECT statements can be very inefficient if done against large tables.

PHP States and PHP Errors frames will show potential issues with PHP. The PHP States frame will show PHP process terminations, starts, and when the service reaches the ready state by node. The PHP Errors frame may help isolate where the issue is with PHP, such as memory size, workers, or the number of servers.

To see latency in transactions, the Transactions – Avg, Max, Min table can be sorted by column to show the longest-running transaction duration. An overloaded cluster will have latent durations in transactions, but it will also show anomalies that might pinpoint a problem with a method or cron.

The Cron error frame will show cron locks, SQL errors that may be associated with cron logs, and shared staging crons that may be running on production environments when there is a dedicated staging environment.

The ElasticSearch Errors frame shows errors that may indicate major problems with Elasticsearch queries, data, or indices.