To me that’s the most interesting part of your experiment. If a little bit of SSD space can offload all the filewalker activity… that’s a huge win! I seriously think your HDDs will last longer if they don’t spend 50% of their life trying to touch the same millions of files over and over…
2 Likes
Likely it is enabled and likely working, BUT
So. You likely need to run it explicitly to reclaim the used space…
Well, what you call “standard settings” is not standard even on Linux not to mention Windows. ZFS is not at the same level of nativeness on Linux compared to EXT4. So anything that requires ZFS is not quite “standard” at least.
ZFS has certain advantages but it’s not a universal fit to be standard. For ex., it has higher RAM demands that might be an issue depending on how many nodes you run on the same host. Of course you can join all your drives using logical volumes and run one node only but it makes your disks interdependent that is not desirable.
1 Like
I think your statement is a bit outdated. It is a single command to install ZFS on a Debian system. It is already part of the standard packages. Also my pi5 doesn’t has a lot of memory to begin with. Sure it would be faster with more RAM but who is claiming that it wouldn’t work with low RAM? I am willing to put that on a test.
And last but not least ext4 also has similar settings that you can try out. So again we are talking about a few standard settings. Just do it instead of complaining.
It might be a single command to install ZFS on a Debian or on any other Linux distributions. It doesn’t make ZFS or many other FSes as native to Linux as EXT4 that requires no installations at all and no kernel updates will ever brake EXT4 support.
I don’t know what other standard settings you’re talking about that would work in all FSes, including NTFS that SNOs use.
As far as I know ext4 calls it an inode cache.
So in other words what you meant by “not super optimized solution” was just add tons of RAM that the OS could use as a disk cache, is that it?
No. I was talking about a metadata cache outside RAM on a SSD.
I am doing it on a pi5 with 8GB RAM total…
To my knowledge, you can’t do it on EXT4 and NTFS, but I might be wrong. If it’s possible I’d love to know how. Quick search gave me this: hard drive - Linux filesystem with metadata on SSD and file data on HDD - Super User . Something that a Linux geek might enjoy researching but not that a regular SNO would be skilled to set up and definitely not something production ready quality.
I am not a Linux geek and haven’t used ZFS before. It doesn’t take long to learn it. We have an awesome community that can help.
What I’m having issues with some posts on the forum is that some people try to lie while telling technically correct things. From our whole discussion it was evident that ZFS is not something that all SNOs would want to go for and I specifically asked multiple times how to do what you claimed possible with EXT4 and NTFS. Yet you’re still responding about ZFS that was not what I asked about.
And all I am reading are more and more excuses. We have a thread here with ext4 inode cache information but sure let’s discard it becaus it doesn’t work ntfs
To tell the truth I have a strong suspicion there are Storj agents among us who troll SNOs and try to manipulate them pretending to be SNOs and using demagogy.
Lol. Now it’s even a conspiracy that prevents you from improving your node. I love it.
2 Likes
1 Like
Conspirancy or not the fact is instead of Storj fixing screwed trash handling you advocate that SNOs “improve” their nodes. Even Storj’s documentation recommend using EXT4 and NTFS and now SNOs need to “improve” by switching to ZFS?
And who benefits from your urging? Storj who can continue mis-handling trash or SNOs who now need to add SSDs to their nodes and more RAM to support ZFS? So you’re a SNO who’s against your own SNO interests and advocate for Storj’s interests? Conspiracy or not, whose intersts you persue that’s who you’re affiliated with.
2 Likes
Ok lets end this discussion. You are not reading. You are just coming up with more and more excuses to not look into ext4 inode caching. You keep claiming that an ext4 inode cache means switching to ZFS.
Let’s end it.
This was my question:
And this was your response:
What am I supposed to look at? If you give a link, I’ll have a look and respond about that link.
I’d be totally OK with ZFS metadata acceleration being a niche thing: used only by SNOs who want to lower disk loads, improve performance, win more uploads, and make more money.
EXT4 and NTFS are fine for common SNOs. 