Set WantMassQuery: 1 on newly created binary caches #810

New issue

Open

opened 2025-04-26 20:43:19 +00:00 by jade · 4 comments

jade commented

2025-04-26 20:43:19 +00:00

Owner

~~Not setting this is a footgun that makes query performance very bad. We should just fix it. See issues like the-distro/infra#201~~

Perhaps we should also remove it altogether and always assume it, but there may be some reason it shouldn't always be done?? Unclear. Surely if that is the case then the people who want slow path querying could set it as some per-substituter setting.

~~Not setting this is a footgun that makes query performance very bad. We should just fix it. See issues like https://git.lix.systems/the-distro/infra/issues/201~~ Perhaps we should also remove it altogether and always assume it, but there may be *some* reason it shouldn't always be done?? Unclear. Surely if that is the case then the people who want slow path querying could set it as some per-substituter setting.

jade added the

Area/store

label

2025-04-26 20:43:19 +00:00

jade self-assigned this

2025-04-26 20:43:19 +00:00

lix-bot commented

2025-04-26 21:54:56 +00:00

Member

This issue was mentioned on Gerrit on the following CLs:

commit message in cl/3063 ("store: WantMassQuery by default")

This issue was mentioned on Gerrit on the following CLs: * commit message in [cl/3063](https://gerrit.lix.systems/c/lix/+/3063) ("store: WantMassQuery by default")

jade commented

2025-04-26 22:19:09 +00:00

Author

Owner

wait a minute. am i stupid? this does seem to only affect querySubstitutablePaths which is basically not used. ughhhh

puck commented

2025-04-26 22:21:55 +00:00

Owner

yeah, it's just like. nix-store and nix-env flags that use it, iirc. I'd be tempted to just remove the distinction; nix is already expensive (in request count) on substituters anyhow

yeah, it's just like. `nix-store` and `nix-env` flags that use it, iirc. I'd be tempted to just remove the distinction; nix is already expensive (in request count) on substituters anyhow

jade commented

2025-04-26 22:24:07 +00:00

Author

Owner

yeah. okay. so we actually mostly do not care and we could delete the whole thing. then my substituter query concurrency problem is in another castle...