Something tells me you measured peak power consumption of system in total, but you won't share those stats with your wife :p

I have one 1500 UPS with the disk shelves on it. One for the Server machine itself

All my systems together are under 1500 watts

Because that is when the breaker trips

You're gonna have a bad day if someone turns on a real 2000 W room heater in your house.

In winter the bedroom is always warm. And in summer I have an exhaust fan

The exhaust fan: newatlas-brightspot.s3.amazonaws.co…monstrator-engine-52566503482-o.jpg

Just picked this one up recently

Are they actually also more efficient than conventional fans?

The amount of times that I bumped into my (luckily plastic) fan blades during summer are immeasurable.

I miss the all metal circle fan used prior. But it was stupid loud

Dyson is fairly quiet for the amount of air it moves

Oh, I wasn't aware of that. I should check those out. I also haven't seen much of any knockoff brands so I'm guessing they got their patents right.

SponiX: Fans are designed to be white noise as well.

Some do a better job than others

Especially acoustically

I wonder if there's a way to exclude some packages from `pkg upgrade`

You can "pkg lock" packages.

Would I need to unlock them before running make -C /usr/ports/$PORTNAME clean build reinstall?

The ports source build is rather outside of the pkg system. I am not sure exactly how they interact. But you can make install and then see the port show up in the pkg list and so on. I don't know if, and I doubt that it does, the pkg lock would block make install. I would think the make install would ignore the pkg lock.

I also think that one can pkg upgrade -f -y and pkg re-install everything overriding what has been installed from source. Switching back and forth like this can fail if needed libraries get swapped out by the other system in incompatible ways. But if that doesn't happen then of course they can work just fine.

Ports hooks into pkg to finally install things, is my understanding.

I don't know if pkg install -f forces pkg lock locks.

is there a secure way to allow the host system to connect to a Unix socket within a jail (e.g., to proxy to a fastcgi application), without the jail being able to replace the socket with a symlink or something like that?

hm, is it possible to create a unix socket beforehand and have a program to bind to it? if that is possible then how about creating the socket on the host and make the directory read-only?

ivy really cool question i'd like to know too

nimaje: i'm not sure this works because most programs will try to unlink() any existing socket before binding it again

possibly the nosymfollow mount option might work

I would think if your progream were written to chck for and re-use the socket if it already exists, this would not be a problem. I've done things like this with a VM using a socket (via VirtualBox) to simulate a serial port in the VM, for example. I think my program opens ot and VB creates it, as I recall

thinking of socket in a jail - if the host accesses it within the host file system (alias'd path) I expect this might work. Have not tried.

ivy: Can you have the socket live in a read-only filesystem? The jail shouldn't be able to do much then.

Might be other ways, but that seems cleanest. Create it, make the filesystem read-only (I'm assuming the socket can survive that), and profit.

ivy: This might also be the sort of thing you could do with, for instance, SELinux, but I've never played with FreeBSD's MAC system.