Comments on: The Cult of ZFS

By: ZFS won’t save you: fancy filesystem fanatics need to get a clue about bit rot (and RAID-5) – Jody Bruchon

Wed, 21 Nov 2018 05:02:10 +0000

[…] the author of this post is incompetent],” would you please feel free to tell that to all the people that say “CRC” when discussing ZFS? Language is made to communicate things and if I […]

By: ZFS won’t save you: fancy filesystem fanatics need to get a clue about bit rot (and RAID-5) | Random musings from a hazardous computer business owner

Thu, 10 Aug 2017 03:55:45 +0000

By: Scott Alan Miller

Scott Alan Miller — Tue, 08 Nov 2016 19:52:49 +0000

ZFS definitely has some nice features and is pretty mature at this point. And it was totally ground breaking in 2005 when it first released, it is great stuff. It’s just important to keep it in perspective that it is a special purpose filesystem, not an end all answer to all storage for all people, which is how it is often treated. And strangely, only treated this way after it was mature and mostly forgotten about and only after BtrFS and ReFS started to appear did people get so excited about ZFS after many years of us preaching the benefits.

We are starting to see a smaller, but still relevant “Cult of ReFS” problem rising now, as well.

By: Peter Molnar

Peter Molnar — Thu, 03 Nov 2016 14:01:13 +0000

Last evening I moved my system to ZFS, and I had good reasons: I wanted compression ( lots of old docs and tonnes of saves web – mostly text – stuff is easier to handle this way than making tarballs out of them ) and I wanted to add my SSD as cache.

The SSD we’re talking about has probably started failing: it’s a mini pci-e Corestore MV card in a ThinkPad X200, which was my root and home system for a long while, while using the spinning disk for storage. In the past 2 weeks I had more system crashes ( Mint 18 ) than in the previous 7-8 years alltogether, so I started to believe it has something to do with the SSD.

Nevertheless I started looking at options and at first, I tried setting up my system with btrfs and add bcache once installed – I failed. Creating the bcache should have been the first step and looking at it’s architecture a bit deeper I’m not convinced it would handle a crash of the cache device well. So when I reached this point I decided to give ZFS a go – the last time I tried it, it burnt me quite hard, having a full-disk encrypted FreeBSD become corrupted with it.

So this time it’s ZFS on LUKS, using the SSD as log and cache device, and so far, I’m really happy and impressed; it was trivial to add and/or remove the device after an install, and so far, it’s happy, although it’s probably early to tell.

What I wanted say is that ZFS has some quite unique features; features, that are not needed for many, but they can come really handy for others, which no other system offers right now at all.

By: Scott Alan Miller

Scott Alan Miller — Wed, 07 Sep 2016 15:07:42 +0000

Thanks, James!

By: James Holden

James Holden — Fri, 26 Aug 2016 22:55:21 +0000

This was a very interesting read. Sadly, this is also the most objective view of ZFS I’ve seen anywhere. Honestly, other than the fact that I mostly work with Linux systems which already default to a very robust filesystem stack, the thing that has made me most wary of ZFS is everyone I know who has told me I should be using it seems to know far less about it and most other filesystem stacks than I do. In no way claim to be an expert in filesystems, so this coupled with the fact that their feelings toward it can usually only be described as “religious epiphany” is especially disturbing to me.

Again, Thank you for your objective view of an excellent file system.

By: Scott Alan Miller

Scott Alan Miller — Tue, 14 Jun 2016 15:04:56 +0000

I’ve been using ZFS for over a decade now. Your response does not appear to be to the article but to an assumption. Did you actually read it? The point of the article was about the near religious fervor with which people promote ZFS as being the one and only system to use, even when not appropriate or useful and without understanding it. Comparing it to NTFS and EXT is telling as those are not its competition… I see that you avoided seeing how it compares to its competition or perhaps you are just not aware of it competition.

The point of the article was NOT to make recommendations, which it did not, other than to avoid choosing ZFS because a zealot has told you that it must be done. IT Professionals need to understand the technologies that they choose and understand the factors involved. ZFS is not a panacea and most of the reasons that it is promoted are often based on incorrect assumptions or outright misinformation. Is ZFS awesome? Of course it is, it has some amazing features and I’ve been promoting it for a very long time.

Hardware RAID is hardly dead or going away. We’ve had a decade of ZFS and hardware RAID hasn’t seen a dent in usage. Why do you feel that now, when nothing new of significance has happened with ZFS do you feel that today ZFS is set to replace hardware RAID when it hasn’t in all of this time? Especially when ZFS’ software side competition has become so much more fierce?

By: Marko O.

Marko O. — Sat, 11 Jun 2016 10:43:52 +0000

Texts like this, promoting hardware RAIDs for no particular reason and bashing ZFS community for no particular reason, really hurt advances in people’s heads because not providing any reason for conclusions.
Ubuntu ships with ZFS now in 2016 and text like this could only hurt people and technology as their intention was just that.
When something is so largely better then NTFS and ext things you grip on, one must find a way of hurting opponent system, even it is miles ahead and multiplatform and in large scale production for more then 10 years already.
Written from a guy using ZFS on laptop for 7 years.
If you want OpenZFS to be better in any way, you can contribute instead of bashing and saying “oh what’s new there”.
Well for hardware RAID news is: “you are out of the job”.

By: H.Trickler

H.Trickler — Wed, 11 Jun 2014 07:08:04 +0000

I wonder how one can publish a long text about ZFS without mentioning the well known design flaw which was botched away by an ugly throttle mechanism.

Imho ZFS was first to implement CRC checks and repair (called scrubbing) for all data and this was the big advantage to all other filesystems.

Ext4 has now implemented this for meta data and if it should become available for all data I could easily imagine that it keeps it’s ‘market dominance’ regardless of BtrFS.