Open PedroMPagani opened 2 months ago
Why not just use the geyser spark extension for profiling? The whole point of extensions is to be able to add functionality that is useful, but not necessarily vital enough to be added directly into Geyser itself.
A JFR profiler records both ways(Memory alloc and CPU usage), has a LOT less overhead than a spark extension, and it's much easier to read/use for investigate? you can see a lot more stuff than a spark profiler, you can see garbage collector, number of active threads, different events like parking, etc, while on spark you're not going to see much.
lot less overhead
Got some data for that?
I'll be honest, so far, spark has been beneficial enough to the point where I don't see the need to add a built-in profiler that barely anyone would use. That would be extension territory. Open to hear what others think, but, this seems like a tad out of scope
What feature do you want to see added?
This will 100% def help improving memory usage and CPU usage of geyser, and possibly even help solving other issues. I think this, along maybe another implementation for a hprof are both benefitial.
Are there any alternatives?
Only through extensions