<colemickens>
petition for a nixos-unstable log|sticker
<samueldr>
a what now?
<colemickens>
thats how petitions work, right, you just write "petition" and then the thing you want
<samueldr>
what do you mean by log|sticker?
<colemickens>
(I'm reading discussions from other rooms and looking at the redbubble shop)
<samueldr>
oh, logo?
<colemickens>
each new nixos release gets a new logo + stickers.
<samueldr>
well, we have to provide incentives to stick to stable
<colemickens>
ha
<colemickens>
I actually really laughed at that :)
<joepie91>
was it someone in here who recommended that I look at nushell?
<samueldr>
I don't know, but if we had pennies still, I'd ask for your thoughts anyway
<joepie91>
lol
<joepie91>
samueldr: so far, it falls into the very very small bucket of "projects where with every documentation page I become more optimistic about the design, instead of more disappointed"
<joepie91>
messed with it a bit today, seems nice
<joepie91>
going to try daily-driving it for a while
<joepie91>
first shell I actually enjoy using, anyway :P
<samueldr>
oh
<samueldr>
that's not a microsoft shell
<joepie91>
nop
<samueldr>
I guess I got my wires mixed by nu-get
<joepie91>
(yeah I was confused at first too :P)
<colemickens>
Have you looked at Rhai too joepie91 ? Same origianl author
<joepie91>
ironically it is basically powershell without the powershell
* colemickens
wants nushell scripting
<samueldr>
joepie91: so it's nuthing?
<samueldr>
powershell - powershell =
<joepie91>
colemickens: nop, not heard of that before
<joepie91>
I did see a different similar project
<joepie91>
samueldr: heh, I'm surprised you didn't catch the meaning :D
<joepie91>
basically it doesn't have the weird powershell syntax and .NET kitchensink attached to it
<colemickens>
I think it's a fairly different thing entirely, but I was also impressed by what I saw of Rhai and was surprised to see it was the same person driving both
<joepie91>
like, not having .NET stapled to it is a big plus in my book
<samueldr>
yeah, I was being playful on the words here
<joepie91>
colemickens: will have to give it a look, thanks
<samueldr>
so, I guess it's a nu start with a nu shell?
<joepie91>
samueldr: I think you have run out of pun quotum
<abathur>
I spent an hour or so playing around with it on a screenshare earlier this year; if nothing else I could see using it to teach shell concepts to beginners, though I guess that kinda sets them up for disappointment, too...
<abathur>
not like teaching someone shell isn't inherently cruel :]
<pie_>
:D
endformationage has quit [Quit: WeeChat 2.9]
<ashkitten>
why are nix eval errors completely useless :<
<ashkitten>
nix: "use --show-trace to show the source of the error" // also nix: [trace contains no mention of the actual source of the error]
<ashkitten>
i know it's caused by setting services.pipewire.pulse.enable = true but i can't figure out what's actually happening lol
<ashkitten>
got it
<ashkitten>
nix traces 😔
<samueldr>
it's because it's a lazy language
<samueldr>
too lazy to keep notes
<samueldr>
too lazy to take some time thinking back about what it just did
<samueldr>
;)
<ashkitten>
samueldr++
<{^_^}>
samueldr's karma got increased to 298
<ashkitten>
git blame-self
<samueldr>
I stopped using git blam on mobile nixos
<samueldr>
blame*
<samueldr>
but blam is a good onomatopeia
<ashkitten>
im annoyed that my local changes broke the build after a clean merge
<samueldr>
shouldn't the CI test against everyone's local changes?
<ashkitten>
yes, obviously :)
<abathur>
git blam just sends a poorly-spelled email to every listserv in your contacts accusing a random person who touched the same code of breaking everything
<ashkitten>
hahaha
cole-h has quit [Ping timeout: 256 seconds]
<samueldr>
>> During his talk, Poettering shared an even more ambitious vision. “I personally think we should go towards systems where etc/ is generally not writeable,” except when an actual reconfiguration is taking place. He argues that creating or updating users just shouldn’t be a configuration change.
<samueldr>
what if it was just not writeable, even when you want to change the configuration?
<ashkitten>
so pipewire is finally useable for me
<ashkitten>
as far as i can tell
<ashkitten>
it's just a drop in replacement for pulseaudio with slightly worse session management now
<ashkitten>
(0.3.16)
MichaelRaskin has quit [Quit: MichaelRaskin]
waleee-cl has quit [Quit: Connection closed for inactivity]
FRidh has joined #nixos-chat
<elvishjerricco>
Apple's acting super proud of their unified memory access (UMA) in the M1. I wonder if that extends to the DRAM cache of the SSD. Like could we get even faster storage by avoiding transferring cached data across pcie?
<gchristensen>
all you need to do is thumbs-up the PR
<gchristensen>
with the github reaction
<red[evilred]>
Will do
cole-h has joined #nixos-chat
__monty__ has joined #nixos-chat
<supersandro2000>
totally read all of that code in 5 seconds
FRidh has quit [Ping timeout: 240 seconds]
rajivr has quit [Quit: Connection closed for inactivity]
<pie_>
<x> >In accordance with their tradition, the Electronic Frontier Foundation has created a new website to tell us how fucked we are by advertisers, while containing almost no advice for fixing any of it
<pie_>
<x> >Having been kicked out of its parents' basement, a detachment of the Rust Evanglism Strike Force wanders into a nearby abbey for shelter, setting up the plot of Season 2, wherein they discover the charity is coming from their actual competitors.
<insep_>
lol it's just a glorified shell script, it just calls to ffmpeg
<cole-h>
ZFS friends: how do I send my entire rpool to another (local) pool? Create a snapshot then `zfs send -wvR rpool@snap | zfs recv bpool/test`?
<insep_>
also i hate how type is repeated in lines 36, 43 and 50, would it compile without them?
<cole-h>
And what would happen if the pool I'm sending from has errors...........
<LinuxHackerman>
cole-h: yep
<cole-h>
LinuxHackerman: And if I ever wanted to "restore" that send... how would I go about doing that? I tried `zfs send -wvR rpool` earlier which gave me an error about unsupported flags or some such. I imagine if I tried the same thing with bpool/test, it wouldn't work.
<sphalerite>
cole-h: oh right, not sure about the -w
<cole-h>
sphalerite: Maybe I can load-key it and `zfs send | zfs recv` it...
<cole-h>
Of course this happens when I really have no time to do anything about it :(
<cole-h>
Good thing I got that 14TB external recently...
FRidh has joined #nixos-chat
<red[evilred]>
always worth testing those things before you hit the two digit terabyte scale ;-)
<cole-h>
My system is only 1TB, so yeah, "easy" to test
<cole-h>
the 14tb disk is where I'm sending to
{^_^} has quit [Excess Flood]
{^_^} has joined #nixos-chat
FRidh has quit [Quit: Konversation terminated!]
<lovesegfault>
,locate ethtool
<{^_^}>
Found in packages: dpdk, ethtool, linuxPackages.dpdk, linuxPackages_4_4.dpdk, linuxPackages_4_9.dpdk, linuxPackages_5_5.dpdk, linuxPackages_4_14.dpdk, linuxPackages_4_19.dpdk, linuxPackages-libre.dpdk, linuxPackages_hardened.dpdk, linuxPackages_xen_dom0.dpdk, linuxPackages_latest-libre.dpdk, linuxPackages_latest_xen_dom0.dpdk, linuxPackages_testing_bcachefs.dpdk, linuxPackages_testing_hardened.dpdk, linuxPackages_latest_xen_dom0_hardened.dpdk
<elvishjerricco>
cole-h: `zfs send -wvR rpool@snap | zfs recv bpool/test` ought to work. Did it not?
<cole-h>
It did
<cole-h>
What I tried was `zfs send -wvR rpool` lol
<cole-h>
(The thing that didn't work, at least)
<elvishjerricco>
cole-h: Ah. Though I would warn you, having an encrypted root dataset means you'll never be able to restore from backup to the same dataset layout; the restored pool will have to contain your datasets in a child dataset
endformationage has joined #nixos-chat
__monty__ has quit [Quit: leaving]
<cole-h>
elvishjerricco: What do you mean?
<cole-h>
(Thought I had sent that earlier, sorry...)
<elvishjerricco>
cole-h: If you make a pool with an encrypted root dataset, then you cannot receive into that dataset
<elvishjerricco>
It'll say you need -F, and then using that will say you cannot use -F on encrypted datasets
<cole-h>
Oh
<cole-h>
That doesn't sound very nice :(
<elvishjerricco>
cole-h: Yea doing a full receive of a dataset with the intention to have it encrypted always requires it be created new
<cole-h>
Oh
<elvishjerricco>
Encryption is... weird with send/recv
<cole-h>
So I just have to `recv` to `rpool2`, `mv rpool rpool.old` and `mv rpool2 rpool`, then?
<cole-h>
(zfs equivalent commands, of course)
<elvishjerricco>
cole-h: what is rpool2?
<cole-h>
Nothing
<cole-h>
Wouldn't receiving into rpool2 create it?
<elvishjerricco>
cole-h: No. The root path component in a dataset name is always a pool. So rpool2 would have to be its own pool that already exists. i.e. it isn't being created new, and needless to say you can't mv datasets between pools
<cole-h>
Ohhhh.....
<elvishjerricco>
Yea the root dataset is special
<elvishjerricco>
I tend to set it to mountpoint=none and never touch it for anything ever, not even encryption. Maybe compression=lz4 though
<cole-h>
Does this mean I should always make the root dataset without encryption, and put encryption on the subsets when I create them?
<elvishjerricco>
Right that's what I tend to prefer
<cole-h>
I mean, `rpool` already is mountpoint=none for me.
<elvishjerricco>
cole-h: I do rpool/root and have all my important settings on rpool/root and all my datasets descendants of rpool/root
<cole-h>
Oh, I see. That's very smart.
* cole-h
updates my setup instructions to do the same
<cole-h>
But then, considering I have a `bpool/backups/rpool/*` (with all the contents of the above), could I just create a new pool on the drive (without encryption), recv those subsets into `newrpool/root` and be fine?
<elvishjerricco>
cole-h: I think so. There's also some quirk with encrypted datasets about incremental sends that I can't remember, so I'm not totally sure
<elvishjerricco>
cole-h: For them to have rpool/root as an encryption root, you'd have to load the key on the backup and send unencrypted to the rpool/root child.
<cole-h>
Just `zfs send bpool/backups/rpool/$smth | zfs recv rpool/root/$smth` after loading the key?
<cole-h>
Or do I just `zfs recv rpool/root` and it creates the necessary descendents automagically?
<elvishjerricco>
cole-h: I think both work
<elvishjerricco>
cole-h: You'll need to make new backups from rpool/root after this though; the key is new so raw sends won't be compatible with the existing backup. No idea if you could continue using the existing backup if you used non-raw sends
<elvishjerricco>
Er, wait, we're making this complicated lol
<cole-h>
haha :D
<elvishjerricco>
cole-h: You could just do `zfs send -Rw bpool/backups/rpool | zfs receive rpool/root` to create root with the existing key without having to load the key on the server. Everything happens at once and you *should* be able to continue using the existing backup
<elvishjerricco>
Let me test that last bit real quick...
<elvishjerricco>
Yea it works. Neat
<cole-h>
Interesting. Something to add to my toolbox hah
<cole-h>
Thanks! elvishjerricco++
<{^_^}>
elvishjerricco's karma got increased to 18
<cole-h>
Kinda wish scrub shows where I have a checksum error...
<cole-h>
I know how non-trivial it is, but a man can dream.
<elvishjerricco>
cole-h: You mean like disk address or which file is broken? It can do the latter. I'd wager zdb can do the former
<cole-h>
Oh? How do I see what had a bad checksum, then? :o
<elvishjerricco>
cole-h: zpool status will tell you to use the -v flag to see which files are affected :)
<cole-h>
Hm
<cole-h>
Then what does a "1" in the CKSUM column mean?
<elvishjerricco>
cole-h: It means one block on the disk failed the checksum verification. This doesn't always result in broken data; if that block was metadata, or you have copies=n, then it can repair from redundant copies
<cole-h>
I have copies=1 :D
<elvishjerricco>
"Applications are unaffected" and "errors: No known data errors" both indicate it was a block with redundancy that it repaired
<elvishjerricco>
So it was probably metadata, which has implicity copies=2 or copies=3 depending on the importance
<elvishjerricco>
Check out `redundant_metadata` in `man zfs`
<cole-h>
And if I clear it and scrub again, what are the chances it goes away?
<elvishjerricco>
cole-h: If there are no known data errors, it would only come back if that sector is just completely bonked in the hardware
<cole-h>
Interesting. So basically, ZFS has already "recovered" from that cksum error due to redundant_metadata=all?
<elvishjerricco>
Oh I got that wrong; it's implicitly copies=2 for all dataset level metadata, but you can set it to just 1 for the least important metadata. Pool level metadata is 3 copies I think
<elvishjerricco>
That's most likely, yea
<cole-h>
I see.
<elvishjerricco>
I would clear and scrub to make sure
<cole-h>
Well, it's certainly worrying considering that earlier, I had read/write numbers in the double digits!! But after a clear and a scrub, they went away, and the cksum went from 8 -> 1 (IIRC)
<elvishjerricco>
Oh gosh. That is worrying. Your drive may be failing