Thank you for this. I think it might be very helpful. You don’t happen to know how to read the data returned from the profiler by any chance?
I can’t get the data from the callgrind as it’s producing an error when i run the command. I can however get the SVG for the cprof. I think it shows me where the leak is, but to me it offers no clues.
Specifically, no, but I would look for any custom integrations that is says are doing bad things and remove them. I think that is the point.
Some of the stuff people wrote are causing thread intrusions into other operations due to being poorly written. These ware causing minor hard to find issues all along, unknown lockups and corruption. Now with the latest python in core, it is outright causing havoc. They need to be removed and written properly.
These are nearly all custom integrations, although if you look at the release notes, a few core integrations were also doing bad things and have been patched in the latest . releases. There are MANY custom integrations doing bad things, however.
You say false alarm, but 99% full on swap is not a normal thing either unless tou have less than 2gb of memory. I would look at that link I send above and see if you have a misbehaving custom integration. Signs are there…
