| 14 May 2026 |
 viraptor | Honestly, I'm finding "yet another service I didn't know existed" every two weeks or so because it starts leaking memory from bigger builds. Life's too short, feedback gets ignored, "sudo kill ..." is the way | 00:33:56 |
| viraptor | * Honestly, I'm finding "yet another service I didn't know existed" every two weeks or so because it starts leaking memory from bigger builds. Life's too short, Apple feedback gets ignored, "sudo kill ..." is the way | 00:34:09 |
 hexa | that's what I do | 00:34:11 |
| hexa | that's what you see in the screenshot up there | 00:34:16 |
| hexa | killall -9 fseventsd | 00:34:23 |
 ttp | high number of read/writes causes it to go up as well iirc
so writing node_modules and nix stores etc can cause it to go haywire
restarting it periodically shouldn't hurt it, but it is a sucky solution for something that should be doing nothing | 00:34:41 |
| hexa | launchd.daemons.fseventsd-reclaim = {
script = ''
killall -9 fseventsd
'';
serviceConfig.StartInterval = 3600;
};
| 00:34:58 |
| hexa | I'm doing. | 00:35:02 |
| hexa | we'll be fine | 00:35:08 |
| hexa | frankly, it started getting worse once I fixed disk space issues | 00:36:17 |
| hexa | which probably increased throughput | 00:36:24 |
| hexa | we also have this | 00:37:07 |
| hexa | # MacOS stores extensive logs in /var/db/uuidtext, which cause high disk usage
# Manually: find /var/db/uuidtext -type f -mtime +7 -delete
launchd.daemons.log-erase = {
script = ''
date
log erase --all
'';
serviceConfig.StartInterval = 3600 * 24;
serviceConfig.StandardErrorPath = "/var/log/uuidtext-gc.log";
serviceConfig.StandardOutPath = "/var/log/uuidtext-gc.log";
};
| 00:37:11 |
| viraptor | My recent "just kill the service" somehow resulted in moving its 20GB memory use under the kernel process with 30k threads instead... So yeah. Quality system. | 00:37:55 |
| hexa | It's fine. | 00:38:30 |
| hexa | macs are an AI platform now | 00:38:34 |
| hexa | with thunderbolt clustering | 00:38:38 |
| hexa | where macs are going you don't need memory anyway | 00:38:58 |
| ttp | 
Download image.png | 00:41:33 |
| hexa | question | 00:59:09 |
| hexa | we have environment.darwinConfig = "/nix/home/darwin-config/macs/nix-darwin.nix"; configured | 00:59:17 |
| hexa | but there is no file at that place on the machines | 00:59:25 |
| hexa | The path of the darwin configuration.nix used to configure the system,
this updates the default darwin-config entry in NIX_PATH. Since this
changes an environment variable it will only apply to new shells.
NOTE: Changing this requires running {command}`darwin-rebuild switch -I darwin-config=/path/to/configuration.nix`
the first time to make darwin-rebuild aware of the custom location.
| 00:59:53 |
| hexa | this is the description for that option | 00:59:57 |
| hexa | is that like /etc/nixos/configuration.nix | 01:00:11 |
| hexa | and since we use darwin-rebuild --flake we can just drop this setting? | 01:00:31 |
 Randy Eckenrode | I don’t know what causes it, but something about Nix causes its memory usage to go out of control. I usually reboot when it gets bad. | 01:01:15 |
| hexa | sigkill works well enough | 01:01:31 |
| hexa | it comes back and I can kill it again the next day | 01:01:43 |
| 19 May 2021 |
|  @grahamc:nixos.org set the history visibility to "world_readable". | 19:22:35 |
