FR pgcollect: add `-a` "all cpus" option - or similar

[GitMensch]

The option to limit the recording (in perf only "by event" or "by user name/id") would be necessary, too (username/id would be fine, "by binary" would be extra cool, "by text match in command line" would be even better).

pgconvert could then use a single "logical" "collection" as root (if necessary) and then the PIDs as logical sub-notes - before adding the appropriate stack; pginfo would provide some details on the PIDs collected.

A feature like this is the one part that is missing most compared to "plain perf" or its rcording + conversion to callgrind format (which misses details but most important "eats CPU like an end-boss").

[ostash]

This is not so easy. perfgrind was created to see callgraphs, and there is no calls between different unrelated processes. You can get take a look on flat profile, but this is already available with standard 'perf'.

Another option is to generated distinct callgraphs for each PID separately, but for system wide profiling this will end with huge number of files, each distinct one for each process running in the system. I do not think anybody is interested in this.

Usually people identify processes they are interested in and then focus on them.

On side note there is "bug" in pgcollect regarding how it handles forks, see here:

perfgrind/TODO

Lines 10 to 13 in 485b3f7

	- create several profiles from collected data. If we collect samples by forking
	new process, we have to enable inherit mode. This means that we will get
	samles not only forked process, but from its children as well. We will need
	to add PERF_SAMPLE_TID to sample_type and change .pgdata format.

[GitMensch]

The issue is that multiple processes will be spawned and end within 3h. The goal is to gather 10 to 20 samples with 3-5 minutes each, but up front only username/-id and binary name (and part of the command line) is known. For perf: record with -a and limit to use, then filter, for example with hotspot. This scenario is the only where perfgrind currently does not fit.

[ostash]

Ok, I see the use case, however an item in TODO list should be implemented first.
I'm wondering whether some workaround is possible in meanwhile, e.g. like discussed here: https://unix.stackexchange.com/questions/688478/how-do-i-run-command-hooks-when-a-specific-process-is-started

[GitMensch]

Ok, I see the use case, however an item in TODO list should be implemented first.

It would be nice to add that issue and the workaround in the TODO :-)

I've implemented the workaround as two cases

• a script placed early in PATH that is named as the executable, which calls pgcollect with the command option using the full path for the command; con: this needs toggling PATH or placing a script in an existing entry - and don't forget to remove that later
• specific in the place where I need that: there's a "runner script" already that executes the executable after setting up some environment variables; I've added an environment variable in front of it, that allows to "only" export the variable to pgcollect; while this works in general there's again the need to set the variable (in this case possibly the login script) and it is harder than necessary to set the a useful output file name... I'm creating "an easy" issue for that next.