-
Beta Was this translation helpful? Give feedback.
Answered by
amotin
Nov 11, 2024
Replies: 2 comments 1 reply
-
I can confirm that. I did an experiment:
zfs create ...
zfs set refquota=1G ...
dd if=/dev/urandom bs=100M of=...
...and got the difference of 40k for 128k recordsize and 28M for 4k
recordsize (created file was 8M larger).
I would imagine that it is hard to do that 100% correctly going through
many abstraction layers in ZFS, thus it is possible to slightly overshoot.
…On Wed, Nov 6, 2024 at 6:00 PM hhb777 ***@***.***> wrote:
Hello, I have a question to ask.
zfs version: 2.1.6
A 200GB file system dataset was created, and after the data was fully
written, it was found that the usedbydataset attribute value of the file
system was larger than the size set by refquota, why? There are no
snapshots or sub datasets available.
As shown in the screenshot below:
image.png (view on web)
<https://github.com/user-attachments/assets/30ce325a-0ea1-403e-931c-811d52624b3f>
Thank You!
—
Reply to this email directly, view it on GitHub
<#16727>, or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ABXQ6HP2E2T2RYRF76CJY3LZ7G5BTAVCNFSM6AAAAABRIENSEWVHI2DSMVQWIX3LMV43ERDJONRXK43TNFXW4OZXGQZTGMJXGI>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
1 reply
-
Thank you for your reply. Reference # 13839 can solve my problem. Thank you! |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
IIRC some overshoot possibility was added there intentionally not too long ago. The problem is that ZFS does not know how much space each specific operation will take until it completes pool-wide transaction group commit. Previously reaching quota triggered extra commit. It resulted in more precise quota behavior, but badly hurt pool performance for other datasets/users. Instead now ZFS allows to go slightly (up to 3%) over the quota. See: #13839 .