Post 9mPpMfxcgsAPt63J4K by trashHeap@fosstodon.org
(DIR) More posts by trashHeap@fosstodon.org
(DIR) Post #9mPog378jMM5FFB5Si by henrikjohansen@bsd.network
2019-08-30T15:06:50Z
0 likes, 1 repeats
I used to run ZFS on Solaris 10 & 11 (back when Sun wasn't owned by the digital equivalent of Mr. Burns from the Simpsons). Are any of the non-solaris implementations any good? I need a *reliable* filesystem for 0.7 PB of initial capacity :flan_confused: Boosts highly appreciated :flan_cheer:
(DIR) Post #9mPog3NnjPgO4uyOQq by ParadeGrotesque@mastodon.sdf.org
2019-08-30T15:08:51Z
0 likes, 0 repeats
@henrikjohansen #FreeBSD is the way to go if you want ZFS.#OpenBSD considers ZFS license as unaccptable.#NetBSD implmentation is unstable as far as I know (updates are welcome).I avoid ZFS like the plague on Linux (XFS is, as far as I know, much more stable) so no idea if #Ubuntu is OK.Make of that what you will...
(DIR) Post #9mPolgRlv5EEgIVzZg by ParadeGrotesque@mastodon.sdf.org
2019-08-30T15:09:54Z
0 likes, 0 repeats
@henrikjohansen I'll add that I know a couple of companies running on FreeBSD + ZFS in production, so, yeah, it is very stable on FreeBSD.
(DIR) Post #9mPosDuqOc7yDkEQcK by j@mathstodon.xyz
2019-08-30T15:11:07Z
0 likes, 0 repeats
@ParadeGrotesque @henrikjohansen I was under the impression that ZFS on Linux has been fairly stable recently, especially given that freebsd is rebasing its ZFS implementation on Linux's, moving from away illumos'
(DIR) Post #9mPp9wNWL3Tj2e5pku by feld@bikeshed.party
2019-08-30T15:14:21.823302Z
0 likes, 0 repeats
@henrikjohansen Yes, Matthew Ahrens has stated that FreeBSD & Solaris derivatives are at equal stability. Linux is quickly catching up, and everyone is moving to the ZFS-on-Linux codebase anyway because all of the upstream development on Solaris derivatives is officially dead now.
(DIR) Post #9mPpMfxcgsAPt63J4K by trashHeap@fosstodon.org
2019-08-30T15:16:22Z
0 likes, 0 repeats
@ParadeGrotesque @henrikjohansen Just going to add a second voice that #FreeBSD really obviously gives a lot of love to their ZFS support and goes out of their way to make sure its well integrated and production ready.Like it's really kinda become THEIR big showcase feature post Solaris.
(DIR) Post #9mPqZ3Q5Obi64dpe6K by emacsen@emacsen.net
2019-08-30T15:30:07Z
0 likes, 0 repeats
@henrikjohansen Am interested in what you find, sadly it looks like the other projects in this space have largely died down (btrfs, I'm looking at you). At the same time there was an argument around that ZFS was poorly optimized for modern storage, that it's still built around a world where cpu is faster than io and we're still using spindles. That's increasingly not the case so I'd be interested in hearing about your experience.
(DIR) Post #9mPqgf4nmt8gtJVJSK by j@mathstodon.xyz
2019-08-30T15:14:26Z
0 likes, 0 repeats
@ParadeGrotesque I don't believe it's purely license issues for OpenBSD (although that it is a part of it), it's also the fact that ZFS is *big*, and would require implementing a sizeable solaris/illumos-translation layer to be properly ported.
(DIR) Post #9mPqgfF5AfMXPCJWTo by ParadeGrotesque@mastodon.sdf.org
2019-08-30T15:31:23Z
0 likes, 0 repeats
@j Yes, ZFS size is also a factor.Which makes sense since OpenBSD Dev like their code base small and auditable.
(DIR) Post #9mPsi7fx8qPLmuhoDg by henrikjohansen@bsd.network
2019-08-30T15:17:59Z
0 likes, 0 repeats
@j @ParadeGrotesque Given that Joyent has commited ritual suicide recently a major contributor to Illumos is gone ... so a switch would make sense.
(DIR) Post #9mPsi7yk0zR8jBUoVM by akpoff@bsd.network
2019-08-30T15:48:15Z
0 likes, 1 repeats
@henrikjohansen @j @ParadeGrotesque ZFS on FreeBSD is reliable and stable. A couple of observations:1) Make sure your system board or add-in card will pass the disks through to the OS (JBOD). ZFS *reallY* wants to own the disks,2) Know also that FreeBSD doesn't support ZFS native encryption, yet. That means your *only* option is GELI. GELI has been around for a long time, so it's reliable, but it means you can't move your disks to another OS.1/
(DIR) Post #9mQ2vR2FTPon8Y7Xou by akpoff@bsd.network
2019-08-30T15:53:23Z
0 likes, 1 repeats
@henrikjohansen @j @ParadeGrotesque 2a) But there's a caveat. The GELI metadata sits at the end of the partition, and is, therefore, not redundant. It can be lost meaning you lose the disk, and the pool. Learn how to back it up if you use GELI.I like encrypted drives, but GELI makes me nervous. I'd rather have the encryption layer inside the pool, not below.2/2https://redmine.ixsystems.com/issues/3206
(DIR) Post #9mQ6EG0YzozgIk00ky by henrikjohansen@bsd.network
2019-08-30T18:13:35Z
0 likes, 0 repeats
@akpoffThis data will be kept for at least 10 years. I don't trust the longevity of Filesystem Encryption mechanisms enough for that usecase; I'd rather do the encryption out-of-band. And since this is Genome data it will need encryption.@j @ParadeGrotesque
(DIR) Post #9mQ6EGNxapi1TIwhE0 by ParadeGrotesque@mastodon.sdf.org
2019-08-30T18:25:31Z
0 likes, 0 repeats
@henrikjohansenJust to be on the same page: is that genome data that also contains PII, such as name or other that permits individual identification? 🤔(GDPR nerd here) @akpoff @j
(DIR) Post #9mQC8pzKuQalsaDxHU by henrikjohansen@bsd.network
2019-08-30T19:31:52Z
0 likes, 0 repeats
@ParadeGrotesqueThe actual datafiles contain no PII ... it's just raw aligned Genome data. The way we store it though let's us map a piece of data to a particular patient in order to fulfill other legislative requirements (like who looked at what from which patient) ... and that method uses encryption. @akpoff @j