Hi,
during last NixCon in London someone (Eelco?) came up with a minimal set of NixOS modules for a working base system. The context of that discussion was how to minimize evaluation time by excluding unnecessary modules.
Does anyone know where to find this list?
TIA
I just stumbled upon [RFC 0022] Minimal module list which links to a “minimal” module list:
The “minimal” configuration replaces
module-list.nixwith a smaller list. (This list can be reduced a bit further by eliminating some unnecessary module dependencies, e.g.hardware/opengl.nixandservices/networking/dnsmasq.nixare unnecessary. On the other hand, some modules that are not needed to evaluate the configuration but that are needed to get a workable system are probably missing.)
I assume this is what you’re looking for.
Yes, exactly. Thank you 
Just a quick follow-up over here. We’ve been investigating making our hydra faster and apart from the issue that we’ve seen our old 15.09 evaluations take substantially longer in the hydra evaluator we tried the reduced module list (i stripped the 15.09 list down to 190 from almost 500 modules) but the effect was negligible. Here’s an instantiation with the full list based on our hydra release expression (using busybox/time):
Command being timed: "nix-instantiate -K -I nixpkgs=. nixos/release-flyingcircus.nix"
User time (seconds): 261.63
System time (seconds): 5.46
Percent of CPU this job got: 171%
Elapsed (wall clock) time (h:mm:ss or m:ss): 2m 35.37s
Average shared text size (kbytes): 0
Average unshared data size (kbytes): 0
Average stack size (kbytes): 0
Average total size (kbytes): 0
Maximum resident set size (kbytes): 8284416
Average resident set size (kbytes): 0
Major (requiring I/O) page faults: 1
Minor (reclaiming a frame) page faults: 518995
Voluntary context switches: 649396
Involuntary context switches: 3127
Swaps: 0
File system inputs: 222448
File system outputs: 0
Socket messages sent: 0
Socket messages received: 0
Signals delivered: 0
Page size (bytes): 4096
Exit status: 0
And here’s the output with the reduced module list:
Command being timed: "nix-instantiate -K -I nixpkgs=. nixos/release-flyingcircus.nix"
User time (seconds): 252.26
System time (seconds): 4.51
Percent of CPU this job got: 182%
Elapsed (wall clock) time (h:mm:ss or m:ss): 2m 20.48s
Average shared text size (kbytes): 0
Average unshared data size (kbytes): 0
Average stack size (kbytes): 0
Average total size (kbytes): 0
Maximum resident set size (kbytes): 6860016
Average resident set size (kbytes): 0
Major (requiring I/O) page faults: 0
Minor (reclaiming a frame) page faults: 428040
Voluntary context switches: 622183
Involuntary context switches: 2339
Swaps: 0
File system inputs: 0
File system outputs: 0
Socket messages sent: 0
Socket messages received: 0
Signals delivered: 0
Page size (bytes): 4096
Exit status: 0
The memory usage went down noticably but that isn’t too much of a concern for us at the moment. On the machines that actually build their system configurations we’ve only seen a reduction from 12 seconds overall time (basically nixos-rebuild switch during a noop evaluation) to 10s and a reduction of memory usage from 2.2g to 2.07g …
So, I’m not sure how @edolstra managed to get a more significant reduction, but his set in RFC 22 was much larger with 50 machines. I guess this means the effect is non-linear and the sizing of a 50/200/500/800 entry module list is such that 800 has become prohibitive whereas 200/500 sit in a kind of middle ground …
At the moment we’re going to try to throw a bit more hardware at it by switching CPUs from an old Xeon X5650 @ 2.67GHz to a newer high clocked E3-1270 v6. We’ll likely keep the old machine due to the high number of cores it offers.
Interesting, thanks for sharing your results
The following nixpkgs issue might also be interesting regarding the evaluation times:
https://github.com/NixOS/nixpkgs/issues/57477
Oh in my last comment on that issue I’ve actually also generated a minimal module list (click on the “> details” link there)