Fedora Cloud 35 Approved To Use Btrfs By Default

[Zan Lynx]

How much CPU does brtfs use these days? That's a metric that is rarely benchmarked, but might be quite important. AFAIR it used to be quite bad. with btrfs.

It is only a problem on SSD, as far as I can tell. When a file gets a lot of fragmentation due to Copy On Write then the CPU usage goes up. That also happens if a lot of snapshots are used.

That said, I don't have a problem with my Fedora NAS. It's using a 250G Samsung NVMe for its root drive with btrfs. I'm running snapper timeline on everything. There are about 50 snapshots.

There is also a small amount of increased CPU and disk usage on the HDD RAID10 because I have it mounted with the "autodefrag" option, so it spends some time rewriting small fragments.

And as always, don't use btrfs for inappropriate types of files and complain about it. If you want to run a database service use XFS or EXT4.

[kloczek]

why RHEL 8 removed ?

Give them a bit more time 😊

[Vistaus]

Three months after: "Fedora disables BTRFS as the default and reverts to EXT4... again"

(just a joke)

“Fedora disables BTRFS as the default and switches to Reiser“

[kloczek]

It is only a problem on SSD, as far as I can tell. When a file gets a lot of fragmentation due to Copy On Write then the CPU usage goes up. That also happens if a lot of snapshots are used..

Bollocks ..
I have aboout 10k snapshots on 2TB (Intel p665) volume used in ~70% from +2 years as development platform (mostly building up to +500 rpm packages 24/7) and I don't see that.
Sys time stays flat (I have access to CPU usage stats for all that time collected in zabbix).
All COW file systems may be hurt only in one scenarion when storage pool is almost full (even ZFS).
btrfs has fixed allocation unit so fragmentation has ZERO impact on non-spindle storage. Nevertjheless even on ZFS it is not an issue.

[kloczek]

How much CPU does brtfs use these days? That's a metric that is rarely benchmarked, but might be quite important. AFAIR it used to be quite bad. with btrfs.

That depends on are you using compression and what kind of files you are storing.

[ezst036]

Three months after: "Fedora disables BTRFS as the default and reverts to EXT4... again"

(just a joke)

If they did ever revert from BTRFS, I would hope they would make XFS the default across their entire stack. I'm not really sure why they didn't do that from the outset since that's the default for the enterprise products.(I think?)

I don't know if individual users necessarily need Stratis, but I think XFS is probably better than EXT4 in most instances so there really wouldn't be any case for RH/Fedora to go all the way back to EXT4 either way.

[rastersoft]

“Fedora disables BTRFS as the default and switches to Reiser“

And finally Fedora will become the killer app

[Zan Lynx]

Bollocks ..
I have aboout 10k snapshots on 2TB (Intel p665) volume used in ~70% from +2 years as development platform (mostly building up to +500 rpm packages 24/7) and I don't see that.
Sys time stays flat (I have access to CPU usage stats for all that time collected in zabbix).
All COW file systems may be hurt only in one scenarion when storage pool is almost full (even ZFS).
btrfs has fixed allocation unit so fragmentation has ZERO impact on non-spindle storage. Nevertjheless even on ZFS it is not an issue.

OK so obviously it doesn't apply to you, but I know what I have seen.

I was running Docker on btrfs on a Dell laptop with NVMe SSD in 2017, and removing a stack of snapshots (the layered Docker image) would hit 85% CPU and less than 10% IO. And it would take minutes sometimes.

Also some of those containers were PostreSQL volumes which seemed to increase in btrfs CPU usage the longer they were in use. The filefrag command would report over 100,000 fragments after a while.

Maybe this fix you're talking about happened after 2017. I haven't tried doing anything like that Docker work recently.

[birdie]

There's a reason not to ever use btrfs: there are basically no tools to recover data from it in case it becomes broken (aside from grep'ping which won't allow you to restore anything but basic text files). Of course, if you back it up each hour - that's not a problem.

[Veto]

And as always, don't use btrfs for inappropriate types of files and complain about it. If you want to run a database service use XFS or EXT4.

Or just keep using BTRFS with selective NOCOW on the relevant files/folders (chattr +C).
Do not use VM images without NOCOW...