Monitoring mainframe applications with CA APM
Article ID: 70702
Updated On:
Products
Issue/Introduction
We have a few applications that are using COBOL, MAINFRAME - DATABASE and MAINFRAME - BATCH as the application type.
Can we monitor these types of mainframe applications using CA APM?
Environment
Resolution
The most APM can do is monitor applications written in Java on Linux running on z/OS. We cannot monitor COBOL.
But you are not completely in the dark.
While bytecode instrumentation is not an option, the EPAgent will give you some visibility.
Essentially, you can either look at a logfile or send messages/synthetics to an application API.
Often, with batch applications, you will have a log entry for the start and finish of the batch. More often, the scheduling of the batch process is being handled by AutoSys or similar. So you have an interface through which to schedule and emit notifications, which are used to synchronize other processes - in this case, writing to a log. This gives an indication of the activity of that process, and you can see if it active (or not) when you are having a performance problem downstream. This is only availability information, but it will help you address and avoid the "blame game".
An application interface is even more useful. You may have an administrative interface for the mainframe app that lets you interrogate its status or other metrics. You simply use the EPAgent to ping that interface periodically (every 1-5 minutes) and then scrape that information and publish to Introscope. If an admin interface is not available, you can use a 'synthetic' - a sham transaction (update, then rollback) just to see if the database is functioning. Or maybe do a query that returns a small table - just make sure it is hitting the database and not the cache. You can do timing around the query and response in the EPAgent - and then publish that as your performance metric.
The EPAgent is quite versatile. You just need to understand how mainframe ops 'knows' that tasks are operating as planned - and then emulate that procedure; again, for Availability status only and maybe a little basic performance info.
Small projects like this are often a gateway to more meaningful mainframe app monitoring via the SMF records, or integrating with CA SYSVIEW or CA UIM. The goal is to have a single version of "the truth", when triaging performance issues in complex environments.
Keep in mind that the spreadsheet survey and sizing tools are designed to address the whole enterprise application pool and find candidate applications that are most readily monitored as well as ones that require a bit more work and strategy. You want to be able to find the best candidate apps for your upcoming deployment as well as start the planning on how you address everything else, in order to devise a plan and motivate the stakeholders.
Feedback
