On 5/31/26 03:08, void wrote: > On Sun, May 31, 2026 at 12:58:10AM +0200, Bugs Beastie wrote: > >> Probably this is the most reasonable direct way to go, to introduce >> "desktop builder(s)"! It will get all "small" packages already built >> from regular builders. It will crunch heavy desktop ports with only >> single poudriere builder, but with all available cores given to >> MAKE_JOBS! >> >> Note that such setup is almost opposite to "regular builders" that >> have a small amount of cores per poudriere builder, but large number >> of poudriere builders! >> >>> Intel(R) Xeon(R) CPU E5-2690 0 @ 2.90GHz / 384GB RAM hw.ncpu=16 SHT=off >>> >> If it scales good further, 96 cores will manage electron in one hour! :) > > What I've done in the past is to have the first build run of all the > massive > ports to be built, with a high -J, tmpfs=all and every port in the list > configured in poudriere.conf to not build these massive ports > simultaneously. > Then have another build run for all the other ports, which completes > relatively quickly because the prerequisites for many things are already > available. > > This at the time resulted in a greater chance of all ports being built. > I think the total time (massive-ports-build+regular-build) may be less > but I'd have to check. > > It seems to me that some ports fight each other when J =>2 and even this > 320GB 20 CPU system will run out of resources (typically swap) if the > build runs with the complete portsbuild list in one go. The above build > never hit swap.
How much swap do you use with that 320 GiBytes of RAM? > > Here's top output during the electron41 build > > last pid: 76928; load averages: 22.02, 21.41, 21.32 up > 8+23:06:29 20:17:14 > 184 threads: 20 running, 164 sleeping > CPU: 89.4% user, 0.0% nice, 10.6% system, 0.0% interrupt, 0.0% idle > Mem: 84G Active, 74G Inact, 29G Wired, 1176M Buf, 188G Free > ARC: 16G Total, 8693M MFU, 4833M MRU, 1368K Anon, 192M Header, 2607M Other > 12G Compressed, 19G Uncompressed, 1.61:1 Ratio > Swap: 4096M Total, 4096M Free > -- === Mark Millard marklmi at yahoo.com
