What does “Lost $N events. Lost $N chunks. – Check IO/CPU overload!” mean?

[RokeJulianLockhart]

[GitMensch]

This is a recording only issue, also output by perf in the text outputs and TUI.

It means perf wanted to record the event but was not able to finish that before the next event came in and buffers were full; you can recheck if you really need the high amount of frequency for the recording (if you do it via Hotspot that is in the "advanced" part of the recording window) and if it is a multi-PID recording (system- or user-wide) check if you need the recording from all of those.

... or record on a faster disk / a physical separate one from the one that the recorded program(s) work on.

[RokeJulianLockhart]

... or record on a faster disk / a physical separate one from the one that the recorded program(s) work on.

I'm using a 2 TiB NVMe SSD, so it'll be that when I recorded with perf record via the CLI, I didn't modify the default frequency. Thanks. Perhaps, I should record via Hotspot's UI in the future.

[GitMensch]

Note: you see the command it is using for the recording as well - so you may use a similar one on the command line.
Just in case: Adding -z during recording can help for the io part - but I guess the main issue is "too much recorded at once" (you forgot to show your recording).

[RokeJulianLockhart]

you forgot to show your recording

@GitMensch, what would have helped to show? I'm a novice at performance profiling.

[GitMensch]

The command line used for recoding :-) [the actual perf invocation]
... and as perf records all sub-processes as well additional information when you are aware that there are a lot of processes spawned from the recorded process.

If you are new: you will like it, even more with hotspot as visualization!