Performance Analysis & Trace Tools

Workload history only goes back a few days when you expected weeks. Which table holds that history, and where do you change how long it's kept?

Response times degraded right after a suspected parameter change. Which transaction shows the change history, and does the same one cover OS parameters?

A slow business process spans several SAP systems, so a single-system ST05 trace only ever sees part of it. What tool traces across system boundaries?

You want to force the classic debugger instead of the new one for a specific analysis. Where do you switch it, and is it a system-wide change?

A developer left a debugging session open and walked away, tying up a work process. What parameter eventually frees it, and what's the related capacity limit?

A program is already running (a background job, or another user's runaway report) and you couldn't start it under SE30 yourself. How do you still measure it?

Two admins try to run an ST05 performance trace on the same instance at the same time. What happens, and how do you check before you start?

You switch on ST05, then ask the end user to reproduce their slow action, but the trace comes back empty. Why?

In the SM50 process overview one work process row is highlighted in yellow. What does that tell you, and where do you change that same setting?

ST05 is active and the right user reproduces the problem, but the trace file still has no matching entries. What did load balancing do to you?

You're about to run an SE30 runtime analysis in production with internal-table operation measurement switched on. Why is that risky, and what's the safe variant?

You want to inspect the live state of the SAP table buffers, not just read their hit-ratio statistics. Which transaction, versus ST02?

A user hits an authorization failure and SU53 doesn't make the cause obvious. Which trace records the actual authorization checks with objects and values, and why not ST05?

The dispatcher isn't listed as a row in the SM50 process overview, yet you need to read its developer trace. How do you get it?

You need trace level 2 or 3 to catch a sporadic error, but you don't want a giant trace file left running after it finally fires. What feature handles that?

Server-side traces show nothing, but users on one PC hit SAP GUI errors that the others don't. Which trace captures the client side, and where do its files land?

Do you want live OS resource consumption or the current OS parameter settings? ST06 or OS07?

A user complains one specific dialog step was slow at 10:42. Why won't ST03N answer that, and what does?

In a summarized SQL trace one statement shows a high Identical %. What does that point you to do, and separately what indexed-read time should make you dig deeper?

You need a breakpoint that still fires when the program runs again later (for example a background re-run). Session breakpoint or external breakpoint, and why?

You made a tuning change and want to prove it helped. What does SAT give you over plain SE30 for that, and what is it NOT for?

Faced with a long ST05 SQL trace, why summarize by SQL statement, and what does the '50% of problems are DB' figure justify in practice?

A program is slow but you have no idea WHICH part. Why is SE30 (ABAP Runtime Analysis) the right entry point, and when is it the wrong one?

You suspect a missing index versus a runaway-growth table. Which transaction for which, ST04 or DB02?

A web request never reaches a work process. Which developer trace file do you open first, and how do you map a symptom to the right dev_* file?

You're tempted to leave rdisp/TRACE at a high level on a live system 'just in case.' Why is that a mistake, and what is the standard live value?

You leave an ST05 trace running on a busy system to catch an intermittent issue. What do rstr/max_filesize_MB (16 MB) and rstr/max_files (10) mean for what you'll actually have when it fires?

What are the ABAP debugger flow-control keys (F5-F8), and which system fields does the program-information view show?

You need to see authorization checks, SQL access and locking for one user reproducing a bug. ST01 logs all of that, so what's the catch versus ST05, and how do you keep it safe?

A user reports one slow transaction. Why is ST05 the right tool, and what is the trap in how you run it?

You're done in the new ABAP debugger. Why doesn't backing out of the window cleanly end it, and what actually terminates it?

When you arm ST05 for an end-to-end slow-transaction analysis, which four trace types can you turn on, and can you run them together?

You need both the SQL trace and the ABAP runtime trace for ONE transaction, correlated. What does ST12 give you that running ST05 and SE30 by hand does not?

Debugging through a SUBMIT or CALL TRANSACTION, the classic debugger drops out but the new one keeps going. What's the mechanism, and why does the classic one die?

An instance crashed and you bounced it before looking at the developer trace. Is the evidence from before the restart gone?

Performance history is missing OS-level CPU/disk data but the in-SAP workload data is fine. Which collector do you check, and how do you tell the two apart?

A runtime-analysis measurement on a busy box comes back empty or truncated. Before re-tracing, what storage limit do you check, and why?

You found a long-running SELECT in ST05 and run 'Enter SQL Statement' to get its EXPLAIN plan. What is the limit on trusting that output?

Why does production refuse nonexclusive debugging with DEBUGGING_NOT_POSSIBLE, and what governs whether you even get exclusive mode?

Reads on a small, rarely-changed config table feel slow. Before touching the database, which SAP-layer monitors do you check, and in what order of suspicion?