|
Tankakern posted:btrfs is fine and of course you should turn off CoW when using it with a db or large vms maybe fsync should be fixed instead
|
#
?
Oct 2, 2021 05:28
|
|
|
|
| # ? Sep 24, 2023 15:11 |
|
|
Mr. Crow posted:In a non-trolly fashion I don't understand how with the backing of Facebook and RHEL the RAID5/6 issues are still a thing. Is it just fundamentally broken? it's in fedora, but that is no guarantee that it's in rhel. I think it'll get testing resources but probably not a lot of engineering resources unless someone pays them to do something facebook talked about what btrfs gets for them: https://engineering.fb.com/2018/10/30/open-source/linux/ Specifically, it's io resource allocation and utilization, which makes a lot of sense, since fair performance utilization can help synchronized tasks become more predictable. I don't think raid 5/6 is necessarily a priority for them, so facebook doesn't put engineering resources towards that use case. Likely because raid 10 is good enough, or they have hardware raid and standalone fibrechannel disk arrays to handle the redundancy.
|
|
#
?
Oct 2, 2021 05:29
|
|
|
RAID 5/6 loving sucks anyway. I don't think anyone really uses it anymore because with disk sizes being what they are today, it takes such a long time to rebuild an array that the chance of a second disk failing while the array is still rebuilding is pretty big.
|
|
#
?
Oct 2, 2021 07:06
|
|
|
Single parity is pretty much obsolete, yes. However, we have half a petabyte of storage in double parity raid, typically 12 disks wide. It can't really be beat if you just need lots of storage. Databases run on the good ol' striped mirrors configuration, though a few also run on double parity raid with a log device. It's fine. Rebuild times are at about 24h during full load, which is pretty close to the average write speed of spinning rust anyway. Of course we don't turn off checksums or COW for our DBs or VMs, because having those guarantees are why we use ZFS for our data in the first place…
|
|
#
?
Oct 2, 2021 08:51
|
|
|
lol at using zfs for other things than a nas
|
|
#
?
Oct 2, 2021 10:43
|
|
|
It's amazing how much time and energy the ZFS zealots spend on anti-btrfs diatribes. Meanwhile btrfs users are just truckin' along, perfectly content that ZFS also exists and is also a valid choice of filesystem.
|
|
#
?
Oct 2, 2021 10:52
|
|
|
oh yeah, if there is something i've thought about btrfs users it is "this sure is a person not at all invested in how their choice of filesystem is perceived in relation to the other options". of course the btrfs users are so uninvested that they wouldn't just post to tell you about it, but you can kind of feel it in their presence; there they are, the enlightened people making good technology choices, while not at all talking about it.
|
|
#
?
Oct 2, 2021 11:24
|
|
|
Do you need a donut to sit on?
|
|
#
?
Oct 2, 2021 11:28
|
|
|
the best people are of course the XFS on LVM users who don’t bother you at all
|
|
#
?
Oct 2, 2021 22:35
|
|
|
is there a compelling reason to use xfs over ext4 these days? either on lvm for my upcoming storage rigg sounds Okay
|
|
#
?
Oct 2, 2021 23:55
|
|
|
psiox posted:is there a compelling reason to use xfs over ext4 these days? i like xfs on linux because nasa uses it, and it reliably works on modest hardware as well. it's needs suiting for me if i'm using linux. i don't know what features ext4 has vs xfs, but you can grow an xfs filesystem and such, it's full featured and robust. i switched to xfs on the rare occasions i use linux, mostly because i had an ext4 filesystem that destroyed itself shortly after install (but well after i'd spent the time to set things up). reinstalled w/ xfs and that pc has worked fine since
|
|
#
?
Oct 3, 2021 00:09
|
|
|
psiox posted:is there a compelling reason to use xfs over ext4 these days? the main advantage is that it's been heavily optimized for multi-threaded, parallel io workloads, particularly when using an nvme drive. it's also very mature and robust
|
|
#
?
Oct 3, 2021 00:53
|
|
|
ext4 is like internet porn reliable, most people use it, nobody talks about it like seriously this has been solved for a long rear end time. Ext4 just works, and I don't ever have to think about whether my filesystem is going to implode and eat ~*mY DaTa*~. It's in every Linux kernel you can find, and even in poo poo like u-boot. It's the filesystem to use if you don't want to give a poo poo about your filesystem, which is the correct point of view.
|
|
#
?
Oct 3, 2021 05:20
|
|
|
Broken Machine posted:i like xfs on linux because nasa uses it, and it reliably works on modest hardware as well. it's needs suiting for me if i'm using linux. i don't know what features ext4 has vs xfs, but you can grow an xfs filesystem and such, it's full featured and robust. i switched to xfs on the rare occasions i use linux, mostly because i had an ext4 filesystem that destroyed itself shortly after install (but well after i'd spent the time to set things up). reinstalled w/ xfs and that pc has worked fine since can XFS shrink yet? I haven’t used it in over a decade.
|
|
#
?
Oct 3, 2021 06:48
|
|
|
DoomTrainPhD posted:can XFS shrink yet? I haven’t used it in over a decade. quick googling suggests that no, as of 2019 this was still not possible imo just use ext4 unless you have some bizarre niche reason to use something more complex in which case you should leave that kind of highly-available filesystem work to an actual storage appliance
|
|
#
?
Oct 3, 2021 08:28
|
|
|
Poopernickel posted:ext4 is like internet porn i agree, and to the point where i indeed do just use ext4 and don't think about it, but i do still hope to see data checksumming/scrubbing roll around in ext4 and xfs, as those are nice not-have-to-worry features. afaik most of the infrastructure is in there since a couple of years back (and was used to build the metadata scrubbing they both do have these days), but still some work to be done.
|
|
#
?
Oct 3, 2021 08:55
|
|
|
I have severe storage admin brain and not having checksumming makes me uneasy. I get an email every time such an error occurs and it makes me extremely paranoid because most of the time dmesg and smart are empty. The disk / controller just returned garbage one day and by the time anyone would notice it would be in our backups as well. I really hope bcachefs continues advancing. It seems to be on a pretty solid basis. Data I don't care about still resides on ext4, because ext4 is the default. XFS is good, but it's also the only file system besides btrfs that has ever completely destroyed itself. One day after a routine reboot $HOME for 40 workstations wouldn't mount, and that was not a fun day.
|
|
#
?
Oct 3, 2021 09:11
|
|
|
Kazinsal posted:quick googling suggests that no, as of 2019 this was still not possible "bizzare niche reason" such as resizing filesystems, which is how the OS updaters on "bizarre niche" systems like iOS and many (most?) Android devices on the planet deal with increases in OS image size? I'm glad you've never had to deal with storage problems, which suggests you're working at a reasonable level of abstraction, but that probably has more to do with there being a team somewhere, whether at your employer or at the vendor whose OS you use, who spend a lot of time to this day working on this "solved problem".
|
|
#
?
Oct 3, 2021 16:03
|
|
|
Spazmo posted:"bizzare niche reason" such as resizing filesystems, which is how the OS updaters on "bizarre niche" systems like iOS and many (most?) Android devices on the planet deal with increases in OS image size? what i learn from this post is that using linux without making complex technical choices yourself, or having a tech support team dedicated to you, is impossible or at least inadvisable
|
|
#
?
Oct 3, 2021 16:08
|
|
|
idk you can just install ubuntu or fedora or something like that and just not worry about the filesystem it uses?
|
|
#
?
Oct 3, 2021 16:38
|
|
|
spankmeister posted:idk you can just install ubuntu or fedora or something like that and just not worry about the filesystem it uses? of course you can, but... *taps thread title*
|
|
#
?
Oct 3, 2021 16:54
|
|
|
Spazmo posted:"bizzare niche reason" such as resizing filesystems, which is how the OS updaters on "bizarre niche" systems like iOS and many (most?) Android devices on the planet deal with increases in OS image size? you -do- know that ext4 is resizable, right? Resize2fs works great. I've shipped it in lots of firmware and it's never been an issue.
|
|
#
?
Oct 3, 2021 17:30
|
|
|
Yeah, it's Xfs that can't be shrunk, ext4 is no problem. It's super useful to shrink disk images so you can use them on a smaller disk as long as there is enough free space.
|
|
#
?
Oct 3, 2021 17:39
|
|
|
Imagine having to unmount to shrink a file system 
|
|
#
?
Oct 3, 2021 18:31
|
|
|
xfs is better than ext4 because it doesnt make/want a lost+found directory off the root of the filesystem
|
|
#
?
Oct 3, 2021 19:42
|
|
|
just lmao imagining having strong feelings about filesystems
|
|
#
?
Oct 3, 2021 20:37
|
|
|
Lysidas posted:xfs is better than ext4 because it doesnt make/want a lost+found directory off the root of the filesystem xfs_repair will create one for you
|
|
#
?
Oct 3, 2021 20:52
|
|
|
down the filesystem hole lies madness, and sometimes murder
|
|
#
?
Oct 3, 2021 21:09
|
|
|
are there any good choices for filesystems with checksumming other than zfs? dangit i feel like that's a must-have for a big nas but i could always go back to running integrit regularly (or whatever the preferred checksum db creation/checking software is these days)
|
|
#
?
Oct 3, 2021 21:14
|
|
|
Yeah, btrfs. Bcachefs is promising, but not ready yet. Those are pretty much your realistic options.
|
|
#
?
Oct 3, 2021 21:16
|
|
|
Well, aside from bcachefs, there is ceph… …but that way lies eldritch horror. With checksums tho
|
|
#
?
Oct 3, 2021 21:41
|
|
|
psiox posted:are there any good choices for filesystems with checksumming other than zfs? dangit i feel like that's a must-have for a big nas but i could always go back to running integrit regularly (or whatever the preferred checksum db creation/checking software is these days) btrfs
|
|
#
?
Oct 3, 2021 23:50
|
|
|
sb hermit posted:down the filesystem hole lies madness, and sometimes murder
|
|
#
?
Oct 4, 2021 00:01
|
|
|
i just have ext4 and its fine. partition is encrypted on the desktop anyway so filesystem selection probably doesn't change anything in terms of performance might try xfs if/when i reformat and drop the encryption since it isn't really needed anymore and the desktop does indeed have an nvme drive everything else is sata (or usb3 for rpis lol) so they'll probably stick with ext4
|
|
#
?
Oct 4, 2021 00:34
|
|
|
encrypt all ur filesystems if for no other reason than you can instantly and securely wipe them if you need to throw them out (or just not wipe them and not care)Lysidas posted:xfs is better than ext4 because it doesnt make/want a lost+found directory off the root of the filesystem
|
|
#
?
Oct 4, 2021 01:05
|
|
|
Reiser4 does checksums. 
|
|
#
?
Oct 4, 2021 01:17
|
|
|
modern processors have built in acceleration for aes encryption/decryption so might as well just turn on encryption unless you can't be bothered to enter the code to decrypt your drives. Also, for enterprises, you can setup network bound disk encryption which makes decryption entirely automated.
|
|
#
?
Oct 4, 2021 01:41
|
|
|
psiox posted:are there any good choices for filesystems with checksumming other than zfs? dangit i feel like that's a must-have for a big nas but i could always go back to running integrit regularly (or whatever the preferred checksum db creation/checking software is these days) turn on integrity for the RAID LV containing the filesystem
|
|
#
?
Oct 4, 2021 05:43
|
|
|
Antigravitas posted:Well, aside from bcachefs, there is ceph… what are the issues with ceph?
|
|
#
?
Oct 4, 2021 10:58
|
|
|
|
| # ? Sep 24, 2023 15:11 |
|
|
It's a distributed network file system not known for its simplicity. It's a Bagger 288 when all you need is a proper shovel. The shovel is ZFS in this analogy, btw.
|
|
#
?
Oct 4, 2021 12:07
|
|