lotuspsychje | good morning | 02:25 |
---|---|---|
wez | Afternoon! | 02:48 |
jailbreak | Morning π π | 10:30 |
wez | Evening! | 12:42 |
wez | .o/ | 13:52 |
lotuspsychje | there's currently a zfs bug on jammy unsolved too | 13:53 |
lotuspsychje | might still wait a bit | 13:53 |
wez | lotuspsychje: Is there a btrfs bug too? | 13:53 |
lotuspsychje | i didnt test btrfs on jammy yet | 13:53 |
linsux | is btrfs a junk or linux equivalent of zfs? | 13:53 |
wez | If not then great! looks like you can work around the zfs issues by using btrfs | 13:53 |
wez | It offers most of what zfs does, depends what you need out of zfs I guess | 13:54 |
murmel | linsux: for now, it depends on what you want to do with it | 13:54 |
linsux | i want to setup home nas for home photo/video, 100G per year | 13:56 |
linsux | i shouldprobably get backup? | 13:56 |
linsux | and use zfs? btrfs? | 13:56 |
murmel | linsux: then it doesn't really matter which you choose. (at raid1 level) | 13:56 |
lotuspsychje | think synology NAS has btrfs by default | 13:56 |
murmel | lotuspsychje: which sadly it not upstream btrfs | 13:57 |
linsux | raid0 is mirroring and raid1 is stripping? | 13:57 |
murmel | additionally they only use it for checksumming the files, raid is by lvm | 13:57 |
murmel | linsux: other way around | 13:57 |
linsux | if i use raid1 i gotto use zfs btrfs not ext4 or xfs right? | 13:57 |
linsux | how likely will i lose data on raid1 | 13:58 |
wez | linsux: raid1 can be done via hardware, it doesn;t need to be done by a file syste, | 13:58 |
wez | m | 13:58 |
murmel | linsux: you can go mdraid/lvm raid1 with ext4, but I would go zfs/btrfs route, just because of checksum/scrubbing feature | 13:58 |
linsux | so, zfs or btrfs? | 13:58 |
murmel | because nobody wants to read every single file to see if corruption happened (over the span of a year) | 13:59 |
wez | raid1 is a 1:1 copy, so depends how many drives are in the array and which ones go down | 13:59 |
wez | raid6 exists BTW | 13:59 |
linsux | and hardware raid is always better than software right? | 13:59 |
wez | linsux: Usually, depends on the raid controller. Some are just firmware for the config and software via drivers for the implementation | 14:00 |
wez | linsux: Others are a full on computer on a card that handles it | 14:00 |
wez | compelte with CPU and RAM | 14:00 |
murmel | linsux: hardware raid is bascially dead | 14:01 |
murmel | as hw raid has quite a few disadvantages | 14:01 |
wez | murmel: Oh? | 14:02 |
murmel | yes? | 14:02 |
wez | murmel: Go on | 14:02 |
wez | I am interested to find out why that is the case | 14:03 |
cbreak | usually, hardware raid is much worse than zfs | 14:03 |
wez | cbreak: hardware raid usually has a battery backup | 14:03 |
cbreak | so? | 14:03 |
murmel | wez: the biggest one is, when your controller dies. you literally need to have the exact same one, and even then it's not guaranteed to work | 14:04 |
cbreak | zfs works without battery | 14:04 |
wez | saves those bits | 14:04 |
wez | precious bits | 14:04 |
murmel | wez: additionally, with the ongoing "convergence" in datacenters, it doesn't matter that the cpu is tilting just because it manages 90+ disks | 14:05 |
wez | .o/ | 14:08 |
cbreak | at work, we've had data loss once because some hardware raid controller failed | 14:09 |
cbreak | and the disks were completely unreadable without it, nor with a replacement | 14:09 |
cbreak | with zfs, this won't happen | 14:09 |
cbreak | not even with btrfs or other lesser software raid | 14:09 |
cbreak | I don't think there's a good reason to waste money on inferior hardware raid solutions nowadays, that we have zfs and CPUs that are more than fast enough to handle it | 14:10 |
murmel | whatever "lesser software raid" means | 14:10 |
murmel | that's another thing, sw raid _is_ cheaper | 14:10 |
cbreak | like mdraid, or the stuff apple / microsoft offers | 14:10 |
cbreak | block level redundancy raids | 14:10 |
JanC | hardware RAID controllers should just use the standard RAID format from linux | 14:10 |
cbreak | JanC: they don't. | 14:10 |
cbreak | nothing's standard about them as far as I can tell | 14:11 |
JanC | I meant "standard" as in a de facto standard :) | 14:11 |
murmel | JanC: how would they lock you in then? ^^ | 14:11 |
JanC | and I mean someone should make such controllers, and it would be a good way to market them: never lose data that way | 14:12 |
cbreak | I use zfs in many usecases, and I've only had data loss so far when I experimented around with alpha versions of zfs or similar | 14:12 |
murmel | cbreak: zfs also has other data loss situations, but very few | 14:12 |
cbreak | JanC: why bother? | 14:12 |
cbreak | murmel: haven't encountered those yet though :) | 14:13 |
murmel | cbreak: I don't know if they fixed it already, but there was an issue with zfs send/receive | 14:13 |
cbreak | the hole birth one? | 14:13 |
cbreak | or the one with encrypted raw sends? | 14:13 |
murmel | can't remember, give me a sec to see on gh | 14:13 |
murmel | but yeah, every fs had it's issues with data loss, the big question is rather which fs do you trust enough with your data | 14:14 |
cbreak | the answer should be: Make backups no matter which one you chose. | 14:15 |
murmel | cbreak: definitely | 14:16 |
cbreak | zfs is super reliable, and defends against corruption caused by hardware more so than any other solution I know of, but it's not immune to problems such as bugs in zfs itself, or the house burning down. | 14:16 |
wez | More than a hardware solution? | 14:16 |
cbreak | heh, sure | 14:17 |
cbreak | hardware raid can't even compete with btrfs | 14:17 |
murmel | i do quite like that btrfs brought a lot of firmware bugs to the surface | 14:18 |
cbreak | https://openzfs.github.io/openzfs-docs/Getting%20Started/Ubuntu/Ubuntu%2022.04%20Root%20on%20ZFS.html <- in case you care about zfs boot on ubuntu, in a more proper way | 14:22 |
cbreak | (I recommend using zbm, and not splitting the root into separate /boot, /usr, ...) | 14:22 |
murmel | cbreak: btw, don't you have to disable the zfs module in the ubuntu kernel when using a different zfs version (can remember doing this on 20.04 | 14:24 |
cbreak | I don't think you have to do that manually, but I've only used the dkms thing once, a long time ago | 14:24 |
cbreak | on ubuntu, I usually stick with the shipped zfs | 14:25 |
murmel | I can remember I looked into it because 20.04 shipped with 0.8.6 and I wanted to see the newer 2.0 release (especially as I was building a new raid system) | 14:31 |
cbreak | 0.8.3. I still use that on a pair of servers I control, works ok. | 14:40 |
murmel | cbreak: yeah I wanted to avoid the pool upgrade later down the road | 14:44 |
=== luis220413_ is now known as luis220413 | ||
jailbreak | Afternoon πβΊοΈ | 17:16 |
=== EriC^^_ is now known as EriC^^ | ||
=== EriC^^ is now known as Guest7921 | ||
=== EriC^ is now known as EriC^^ |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!