These 'horns5' work units take too much memory for what I have
available so I've taken to just deleting them. They typically suck ~ 4gigs of
RAM on my boxes, and with 4 cores I end up going to swap memory and then I have paused
jobs while the box waits to free up memory. I only have a few boxes with 16g of RAM most
are at 8 and I think I might still have one or two
with 4g which of course are barely adequate to run 4 cores and Rosetta.
So I've upgraded most of those but they are older and DDR2 ram is needed.

This morning I see another 'hornsx' flavor, maybe '1' or something I didn't
write it down. Maybe they won't be so memory intensive.

Anyhow I'm not trying to gum up stuff but I like to keep these 4 core cpus busy and not running
at 50% while they wait for memory to free up.
So that has been my 'fix'.

/M

---

Anyhow I'm not trying to gum up stuff but I like to keep these 4 core cpus busy and not running at 50% while they wait for memory to free up.
So that has been my 'fix'.

Boinc server scheduler can use AppPlan

Plan classes are policies that allow the BOINC scheduler to decide:

- whether an application should run on a particular host;
- what resources it will use (# of CPUs and GPUs, optionally memory usage)
- how fast it is expected to run.

---

Yup same for me, had to cancel all horns5* because the memory usage was far too much considering I run a plethora of 2GB SBCs.

Now this morning I find the 'True_Relax_SAVE_ALLxxxx' are sucking all available RAM and running the swap file
up to the point where the box nearly dies.

watch those, I'm wondering if they have a memory leak. Seems like it takes them 10 mins. or so before
they really start 'sucking' it all up fast.

---