202: document the worker/memory tradeoff r=Mic92 a=Mic92



Co-authored-by: Jörg Thalheim <joerg@thalheim.io>
This commit is contained in:
bors[bot] 2023-04-13 21:25:47 +00:00 committed by GitHub
commit a5bb26e9ea
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
2 changed files with 36 additions and 5 deletions

View file

@ -91,3 +91,32 @@ latest release branch.
- [colmena](https://github.com/zhaofengli/colmena) - A simple, stateless NixOS deployment tool - [colmena](https://github.com/zhaofengli/colmena) - A simple, stateless NixOS deployment tool
- [robotnix](https://github.com/danielfullmer/robotnix) - Build Android (AOSP) using Nix, used in their [CI](https://github.com/danielfullmer/robotnix/blob/38b80700ee4265c306dcfdcce45056e32ab2973f/.github/workflows/instantiate.yml#L18) - [robotnix](https://github.com/danielfullmer/robotnix) - Build Android (AOSP) using Nix, used in their [CI](https://github.com/danielfullmer/robotnix/blob/38b80700ee4265c306dcfdcce45056e32ab2973f/.github/workflows/instantiate.yml#L18)
## FAQ
### nix-eval-jobs consumes too much memory / is too slow
By default, nix-eval-jobs spawns as many worker processes as there are
hardware threads in the system and limits the memory usage for each worker to
4GB.
However, keep in mind that each worker process may need to re-evaluate shared
dependencies of the attributes, which can introduce some overhead for each
evaluation or cause workers to exceed their memory limit. If you encounter
these situations, you can tune the following options:
`--workers`: This option allows you to set the number of evaluation workers that
nix-eval-jobs should spawn. You can increase or decrease this number to
optimize the evaluation speed and memory usage. For example, if you have a
system with many CPU cores but limited memory, you may want to reduce the
number of workers to avoid exceeding the memory limit.
`--max-memory-size`: This option allows you to adjust the memory limit for each
worker process. By default, it's set to 4GiB, but you can increase or decrease
this value as needed. For example, if you have a system with a lot of memory
and want to speed up the evaluation, you may want to increase the memory limit
to allow workers to cache more data in memory before getting restarted by
nix-eval-jobs.
Overall, tuning these options can help you optimize the performance and memory
usage of nix-eval-jobs to better fit your system and evaluation needs.

View file

@ -90,8 +90,10 @@ struct MyArgs : MixEvalArgs, MixCommonArgs {
.labels = {"workers"}, .labels = {"workers"},
.handler = {[=](std::string s) { nrWorkers = std::stoi(s); }}}); .handler = {[=](std::string s) { nrWorkers = std::stoi(s); }}});
addFlag({.longName = "max-memory-size", addFlag(
.description = "maximum evaluation memory size", {.longName = "max-memory-size",
.description =
"maximum evaluation memory size (4GiB per worker by default)",
.labels = {"size"}, .labels = {"size"},
.handler = { .handler = {
[=](std::string s) { maxMemorySize = std::stoi(s); }}}); [=](std::string s) { maxMemorySize = std::stoi(s); }}});