ext4 slow delete on huge filesystem

[chaostheo]

[Administrator note: this post, and the first 2 responses, were originally attached to the topic Any advice on reflinks... or XFS in general?, but that topic is about XFS reflinks, and this post appears to be about ext4 deletion performance, so I split it out. The thread title ext4 slow delete on huge filesystem was set by me, and may not reflect what OP would have picked. -Hu]

You may have much time in your live when sticking with ext4 ..., running rm for >8h and later a find too >3h now while both seem to take no end on ext4@raid5 but for each user it depends what is needed or wished.
root@pnode02:~# ps -ef|grep 5/r23
root 3660513 3660254 11 10:02 pts/0 00:57:36 rm -rf 5/r23 5/r24 5/r25 5/r26 5/r27 5/r28 5/r29 5/r3 5/r30 5/r31 5/r32 5/r33 5/r34 5/r35 5/r36 5/r37 5/r38 5/r39 5/r4 5/r40 5/r41 5/r42 5/r43 5/r44 5/r45 5/r46 5/r47 5/r48 5/r49 5/r5 5/r50 5/r51 5/r52 5/r53 5/r54 5/r55 5/r56 5/r57 5/r58 5/r59 5/r6 5/r60 5/r61 5/r62 5/r63 5/r7 5/r8 5/r9
root 3909600 3909446 0 18:34 pts/3 00:00:00 grep 5/r23

root@pnode02:~# ps -ef|grep find
root 3820484 3660254 9 15:31 pts/0 00:16:56 find /usr2 -ls
root 3909653 3909446 0 18:34 pts/3 00:00:00 grep find

root@pnode02:~# date
Sun Mar 2 06:37:26 PM CET 2025

root@pnode02:~#: df -h /usr2
Filesystem Size Used Avail Use% Mounted on
/dev/sdb 5.5T 3.1T 2.2T 59% /usr2

root@pnode02:~#: df -hi /usr2
Filesystem Inodes IUsed IFree IUse% Mounted on
/dev/sdb 175M 106M 70M 61% /usr2

[szatox]

So, chaostheo, what's the idea there? Looks like you wanted to show us something, but all I can tell is that your calendar doesn't get ahead if its time.
_________________
Make Computing Fun Again

[chaostheo]

I know when running on single nvme it's faster ... but you cannot compare a multiple of 100's TB hdd raid array with a single nvme.
Even in rh/rocky9 you cannot create ext4 larger than 50TB ... so eg taking 24x 20TB hdd raid6 with ext4 is totally impossible without couple of partitions ...