hello there,
Right now i am running truenas scale with nfs share export and this is insanely slow compared to smb.
I know this is an issue with NFS and ZFS each featuring their own sync write implementation causing slowdown. at the moment i have a separated dataset for nfs where sync write is disable.
I trust zfs sync more than nfs sync so is there a way to export nfs share with async write enable?
Is it? I never had problems with Linux & NFS & ZFS.
Async is the default.
As long as you don’t force ZFS to be sync, yes async is “enabled”.
As long as you don’t force ZFS to be async, yes async is enabled but the client decides if it wants to use sync.
As long as you don’t specify something else for the NFS mount, it uses the async default.
Async is the default.
As long as you don’t force ZFS to be sync, yes async is “enabled”.
from what i understand NFS is sync by default on a normal linux distro.
As for zfs sync “Standard” from What i understand is that it sync some operation under the hood. but not all of the operation is force sync.
As long as you don’t specify something else for the NFS mount, it uses the async default.
normally i use default and fstab.
AIUI, the resulting sync/async for underlying drives can be described by the following table:
| Client issues… \ Dataset | sync=standard | sync=disabled | sync=always |
|---|---|---|---|
| sync write | sync | async | sync |
| async write | async | async | sync |
Soooo, I don’t really understand what you want to achieve. I never used NFS, but I’m 99% sure you can mount it with async writes, even if it is sync by default.
hmm… sorry looks like I was wrong.
Looks like fstab uses sync by default, also looks like TrueNAS honors sync request by default.
Because of the good performance I got, I assumed that it was async. Maybe it still is, because the applications don’t ask for sync?
Sorry again, for causing confusion without helping you.
