Consider an order request that is waiting on an inventory service. Inventory, in turn, may be waiting on another dependency. Looking at the services separately makes it hard to see where to investigate first.
JENNIFER MSA monitoring keeps those service relationships in view as you move into individual transactions. The aim is to follow the affected request far enough to understand what each application was doing.
Follow the service relationships
JENNIFER’s MSA map groups connected services and lets you filter the view to a selected cluster. In this view, a cluster is a grouping within the service map. Use it to focus on the call relationships relevant to the problem.
From the domain view, you can move through instances and applications to individual transactions and their Call Chain. Nodes and connections show information about errors, call volume and response time. Dynamic layout and group movement help keep the map readable as you explore.
Look for a delay concentrated in one application or a dependency shared by several affected paths. Either can give you somewhere to start. You still need to inspect the transactions to test the explanation.
Select the affected requests
In X-View, select the incident window and the transactions you want to examine. A few unusually slow requests suggest a different scope from a slowdown affecting the application throughout the incident.
X-View analysis provides application statistics alongside transaction scatter plots, a chart-focused view and a table of transaction fields. You can also examine response-time components such as SQL time, fetch time and external-call time.
Narrow the period, target and response-time range until the result is manageable. Where the required fields have been collected, grouping and filtering can isolate an application, transactions with errors or a particular request group.
Read profiles in context
JENNIFER’s transaction analysis combines profile inspection with related-transaction grouping and topology. You can examine the recorded work without losing track of how the selected transaction relates to the rest of the request.
Suppose the inventory call accounts for much of the order transaction’s duration. If the related transaction is available, compare its timing and recorded work with the caller’s. The order service may simply be waiting. Its duration alone does not establish where the fault began.
Check the explanation against other affected requests, using the same incident window as you move between views. In the incident record, distinguish the observed delay from your hypothesis about its cause. Include the affected request and time window before assigning a cause or an owner.
Check the installed capabilities
JENNIFER introduced MSA analysis in 5.6.4.12 and the MSA real-time dashboard in 5.6.5.
Before adopting this workflow, check the capabilities and instrumentation supported by your installed server and agents. The service map can guide the investigation, but the next engineering decision should rest on what the affected transactions show.