Work Processes & SAP Services

rdisp/ROLL_MAXFS and rdisp/PG_MAXFS both size an area made of a buffer plus a file in 8KB blocks. What does each one actually hold?

A dialog work process quietly restarts itself right after finishing a step, and nobody killed it. What setting causes this, and why is it a feature rather than a fault?

In SM50 a dialog work process shows the stop reason PRIV. What does that tell you about who can use that work process right now?

A colleague panics that many work processes in SM50 show status Stops and assumes they've crashed. Are they right?

A work process failed and the dispatcher would normally restart it, but SM50 shows its restart value set to no. What situation makes the system deliberately stop retrying?

You have a stuck work process and are about to kill it in SM50. Which two process types must you never cancel manually, and what's the risk?

In the SM50 semaphore column one work process shows a red semaphore number and another shows green. What's the difference?

You need to see work-process load across all instances of the system at once, not just the one you're logged into. Which transaction, and how does it differ from SM50 and SM51?

With dynamic work processes enabled, what caps how many extra work processes the dispatcher will spin up under load, and what becomes of them afterward?

Parallelizing load with cascades of RFCs can deadlock a server. What's the mechanism, and how do dynamic work processes rescue it?

In the default setting the configurable work-process count for an operation-mode switch is rdisp/wp_max_no minus 2. What are those 2 reserved for, and can you reclaim them?

What is the hard upper limit on the number of work processes on a single instance, and what does that ceiling include?

At instance startup, how many work processes come up, and what single constraint must that number satisfy?

rdisp/wp_no_enq maxes out at 5, while dialog and batch counts can climb to 600. Why is the enqueue ceiling so tiny?

In a typical OLTP SAP system, which work-process type normally makes up the largest share, and why?

A dialog work process shows the SM50 stop reason VB. What is it waiting for, and why does that tie up the dialog work process?

When you click in the SAP GUI, which component receives that request first, and how does it decide which work process runs it?

In the SM50 process overview you see DIA, UPD, UPD2, ENQ, BTC and SPO, but never the message server or the gateway. Why don't those two appear as work-process types?

Two instances both offer dialog, update and spool services, but only one also runs the message server. What is that one called, and what are the others called, even though they run several services?

How many gateway processes does an instance need, and must an administrator configure one before RFC and CPI-C will work?

An operation-mode switch and dynamic work processes both adjust work processes during running operation. What can a dynamic work process do that an operation-mode switch fundamentally cannot?

What are reserved (restricted) work processes, what type are they always, and when does the system actually use them?

When VMC is active, why must the Java VM pool contain more VMs than the number of configured work processes?

How does the VMC avoid every Java VM loading the same classes into memory separately, and which ABAP mechanism is it analogous to?

Which transactions administer the VM Container, and what does each one show?

Why does every SAP instance require at least two dialog work processes, even a small one that seems like it could get by with one?

Between two dialog steps of the same user, where does that user's context go while the work process is off serving someone else?

A dialog work process dies vs the dispatcher dies — which one takes the whole instance down, and why the difference?

Why can't you spread enqueue work processes across several instances for redundancy?

Why does SAP run certain Java components (e.g. the Internet Pricing Configurator) inside a VM Container in the ABAP work process, instead of a separate Java stack?

A business transaction can span several dialog steps and many database operations. How does SAP still guarantee all-or-nothing across the whole thing?

Updates split into V1 and V2. Your system has no V2 work process configured — does V2 work just silently not happen?

Your system has no separate (A)SCS instance. Where do the message server and enqueue process run then — and why is that worse for availability?

You run 4 app server instances for redundancy. A colleague says 'so we're safe if one node dies.' What single component still takes the whole system down — and why doesn't adding app servers help?

ENSA1 vs ENSA2 for lock high availability — what can ENSA2 do on failover that the classic model can't, and why does it matter for modern clusters?

True or false: the dispatcher is a different program from the work processes. Why does the answer surprise people?

What are the minimum work-process count rules for an SAP system and for each instance?

If the enqueue server crashes with no replication in place, what happens to in-flight work — and what prevents it?

Why does SAP carve the message server and enqueue out into their own (A)SCS instance instead of leaving them on a normal app server?

TemSe can hold spool data either in the database or in the file system. What do you trade off when you choose one over the other?

A user runs a 5-step transaction. Are they tied to one dialog work process for all 5 steps? Why does this let a handful of WPs serve hundreds of users?

Asynchronous (update-task) database changes — is whether to use them an admin/runtime setting, or decided somewhere else?