Process Snapshot metrics

[garysavage]

When a Process Instance Snapshot is triggered, are the processes captured based on top consumers of the metric that caused the trigger, or is it always based on CPU? I ask because CPU is not always the correct criteria if it’s another resource issue. Also, can this be enhanced to allow the selection of which metric is the core criteria? And, possibly to filter on specific process executable string?   

Can some metric of disk I/O be added to this process snapshot (Network in/out may not represent a disk I/O issue)?

Can the start time of a process be added to the columns or offered as an option?

Can the Parent PID be added to the columns or offered as an option?

Can the Number of processes captured be increased beyond 100?

Will AIX have this process snapshot capability at some point?

We often require process details to resolve diagnostic issues and without the capabilities above, it forces us to transition to alternative tools. It would be great if we could stay within DT to resolve....

Thanks!

[mateusz_marek · Accepted solution]

Hi @garysavage 

When a Process Instance Snapshot is triggered, are the processes captured based on top consumers of the metric that caused the trigger, or is it always based on CPU?   I ask because CPU is not always the correct criteria if it’s another resource issue.

---

---

Top processes are calculated based on "combined resource usage". Firstly, it is calculated how much of each type of monitored resource (CPU, memory, incoming network traffic, outgoing network traffic) is consumed by all the processes in total. Then for each process's resource it's determined how much of that total resource usage is caused by that individual process. "Combined resource usage" of the process is the average from all the percentages that this process's resources occupy. A process will be added to top list only if at least one of resources % of total usage is above minimal, 5% threshold.

---

Also, can this be enhanced to allow selection of which metric is the core criteria?  And, possibly to filter on specific process executable string?   Can some metric of disk I/O be added to this process snapshot (Network in/out may not represent a disk I/O issue)?Can the start time of a process be added to the columns or offered as an option?Can the Parent PID be added to the columns or offered as an option?Can the Number of processes captured be increased beyond 100?

---

For enhancements please create post on Dynatrace product ideas section.

---

Will AIX have this process snapshot capability at some point?

---

Process instance snapshot capability for AIX will be visible on new host screen, with 1.249 version of Dynatrace.

---

Best Regards

Mateusz

[garysavage]

Mateusz, thank you for this information.  Very helpful!    There are always exceptions when data is filtered and calculated before being presented.  Example: if a process spawn storm is consuming the majority of CPU, you might have thousands of processes using very little CPU per process, and with a 5% threshold it sounds like they might not even make the list?

Thanks again!

G