DX Application Performance Management

Hi,

what is the max number of components that can be displayed in 1 Transaction Tracer Viewer (TTV)?

because the cross-process TT worked fine = displaying 2 processes. however we needed to enable the monitoring for all the methods (test env) for the 2 process(app) and the TTV just stopped displaying the second process after that. it displays the components of the 1st process and just a line (with some random method 1-2sec) instead of the 2nd process.

if i open the TTV under the second agent, i can see the TT for the 2nd process, but we need both processes together in 1 view. 

any idea what the reason could be and how to solve it?

thank you,

Dear Team:

Can someone provide next steps for Stefan and how to link the two processes in one view?

Thanks

Hal German

Please help us by always stating your APM version and agent types/app containers.

Hiko, maybe more clearly the issue is this:

we have an issue to display correctly the cross-process TT. the TT view displays correctly the 1st process and all it's components, but instead of the second process (and it's components) just a random (1-2s) method from the 2nd process is displayed so it looks like a line = nothing else.

the issue is active, when we enable very broad instrumentation = thousands of methods. if we restrict the instrumentation to tens of methods, the TT view displays correctly the cross process TT and both processes in the TT view.

details:
1. when we start monitoring the second (.NET) app with pbd:
SetFlag: MercuryCustom
TurnOn: MercuryCustom
IdentifyClassAs: Extract.IS.Class2 MCustom
IdentifyClassAs: Extract.IS.Class2 MCustom
TraceAllMethodsIfFlagged: MCustom BlamePointTracer "WebServices|Methods|{classname}|{method}"
then the cross-process TT and both processes are displayed correctly.

2. however, if we expand the instrumentation and start monitoring the second (.NET) app with pbd:
SetFlag: MercuryCustom
TurnOn: MercuryCustom
IdentifyMatchingClassesAs: Extract.IS.* MCustom
TraceAllMethodsIfFlagged: MCustom BlamePointTracer "WebServices|Methods|{classname}|{method}"
= the whole namespace that covers thousands of methods, then instead of the second process (and it's components) in the cross-process TT view, we see just 1 random (1-2s long) method.
if we go to 2nd agent, then there the TT view displays correctly the 2nd process. so the issue is "just" with the cross-process TT view that should be displayed under the first agent.

3. when we change it back to
IdentifyClassAs: Extract.IS.Class2 MCustom
IdentifyClassAs: Extract.IS.Class2 MCustom
TraceAllMethodsIfFlagged: MCustom BlamePointTracer "WebServices|Methods|{classname}|{method}"
then the cross-process TT and both processes are displayed correctly (both on the first and the 2nd agent as well).

it's APM 10.5.2 and .NET 4.5.2.

You have only 2 processes, with the 1st process that calls the 2nd process, so selecting 1st agent  you should be able to see both processes in the same TT Viewer. Selecting the 2nd agent you should be able to see only the second process TT in the viewer. Certainly you should be able to see at least 2 TT in the caller agent TT Viewer,  there isn't limitation with only 2 processes involved in CrossTT.

The 2nd process viewed as a line, could be a problem of time scale, have you tried to work with zoom cursor?

Hi Marco, you are correct with your description. 1st calling 2nd, etc , but when we use instrumentation

IdentifyMatchingClassesAs: Extract.IS.* MCustom
TraceAllMethodsIfFlagged: MCustom BlamePointTracer "WebServices|Methods|{classname}|{method}"

that covers 10thousands of methods

then the second process just disappear from the TTV.

if we change the instrumantation to more specific (=identifying just couple of classes), then the TTV displays both processes. so it's not a time scale issue, but probably some overloading. any idea in this "area"? thanks

Hi Stefan, thank You for these further details. 
I have never encountered this kind of problem, we have always worked on a limited set of methods and components. It seems to be a problem of overload and the issue is very interesting.

more precisely:

we have an issue to display correctly the cross-process TT. the TT view displays correctly the 1st process and all it's components, but instead of the second process (and it's components) just a random (1-2s) method from the 2nd process is displayed. so it looks like a line.

the issue is active, when we enable very broad instrumentation = thousands of methods. if we restrict the instrumentation to tens of methods, the TT view displays correctly the cross process TT and both processes in the TT view.

details:
1. when we start monitoring the second (.NET) app with pbd:
SetFlag: MercuryCustom
TurnOn: MercuryCustom
IdentifyClassAs: Extract.IS.Class2 MCustom
IdentifyClassAs: Extract.IS.Class2 MCustom
TraceAllMethodsIfFlagged: MCustom BlamePointTracer "WebServices|Methods|{classname}|{method}"
then the cross-process TT and both processes are displayed correctly.

2. however, if we expand the instrumentation and start monitoring the second (.NET) app with pbd:
SetFlag: MercuryCustom
TurnOn: MercuryCustom
IdentifyMatchingClassesAs: Extract.IS.* MCustom
TraceAllMethodsIfFlagged: MCustom BlamePointTracer "WebServices|Methods|{classname}|{method}"
= the whole namespace that covers thousands of methods, then instead of the second process (and it's components) in the cross-process TT view, we see just 1 random (1-2s long) method.
if we go to 2nd agent, then there the TT view displays correctly the 2nd process. so the issue is "just" with the cross-process TT view that should be displayed under the first agent.

3. when we change it back to 
IdentifyClassAs: Extract.IS.Class2 MCustom
IdentifyClassAs: Extract.IS.Class2 MCustom
TraceAllMethodsIfFlagged: MCustom BlamePointTracer "WebServices|Methods|{classname}|{method}"
then the cross-process TT and both processes are displayed correctly (both on the first and the 2nd agent as well).

it's APM 10.5.2 and .NET 4.5.2.

if we are on the second agent = second process TT view: is there please any possibility to get redirected to the first process TT view? to see/know how the first (caller) process looked like? and ideally the whole cross-process TT view.

it could be a workaround, but i couldnt find any redirection from the 2nd process.  i know it's possible via historical event viewer (via the cor proc id), but maybe it;s possible with some easier way.  thank you

I'm going to make the same comment here that I did on Yammer: why aren't you using BackendMarker as your tracer type?

Custom Instrumentation with CA APM 

TT component clamping hopefully isn't the issue, but you should run the agent logging in DEBUG mode to be sure. Do the same at the EM.