[Bug 276299] Write performance to NFS share is ~4x slower than on 13.2
- Reply: bugzilla-noreply_a_freebsd.org: "[Bug 276299] Write performance to NFS share is ~4x slower than on 13.2"
- Reply: bugzilla-noreply_a_freebsd.org: "[Bug 276299] Write performance to NFS share is ~4x slower than on 13.2"
- Reply: bugzilla-noreply_a_freebsd.org: "[Bug 276299] Write performance to NFS share is ~4x slower than on 13.2"
- Reply: bugzilla-noreply_a_freebsd.org: "[Bug 276299] Write performance to NFS share is ~4x slower than on 13.2"
- Reply: bugzilla-noreply_a_freebsd.org: "[Bug 276299] Write performance to NFS share is ~4x slower than on 13.2"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Sat, 13 Jan 2024 15:41:02 UTC
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=276299
Bug ID: 276299
Summary: Write performance to NFS share is ~4x slower than on
13.2
Product: Base System
Version: 14.0-RELEASE
Hardware: arm64
OS: Any
Status: New
Severity: Affects Only Me
Priority: ---
Component: kern
Assignee: bugs@FreeBSD.org
Reporter: dmilith@me.com
I've upgraded my aarch64 box (RockPro64 with 4G RAM) which serves a ZFS pool
over NFS.
Before the upgrade (on FreeBSD 13.2) my upload speed to the share was like
50-70MiB/s. After the upgrade to FreeBSD 14.0, it struggles to get even 10MiB/s
write.
I've tried switching from NFSd v3 to v4. It's even worse (6MiB/s write).
The disk is WD Gold, 10TiB. So I'm quite sure it's not about the disk speed or
the network speed (my router has 1G ports).
My /etc/rc.conf:
# NFS
hostid_enable="YES"
nfscbd_enable="YES"
rpcbind_enable="YES"
nfs_server_enable="YES"
nfsv4_server_only="NO"
nfsv4_server_enable="YES"
nfsuserd_enable="YES"
mountd_enable="YES"
mountd_flags="-r"
rpc_lockd_enable="YES"
rpc_statd_enable="YES"
I've bumped /etc/sysctl.conf settings to huge values:
net.inet.raw.maxdgram=262144
net.inet.raw.recvspace=1048576
net.inet.tcp.sendspace=1048576
vfs.nfsd.srvmaxio=1048576
vfs.nfsd.maxthreads=128
net.inet.tcp.rfc1323=1
net.inet.tcp.sendbuf_max=16777216
net.inet.tcp.recvbuf_max=16777216
But this improved nothing. Maybe it's even worse…
My NFS client (macOS 13+) uses these options:
mount_nfs \
-o \
rw,vers=4,deadtimeout=0,readahead=6,noatime,sync,async,hard,bg,intr,inet,tcp,nfc,rsize=1048576,wsize=1048576,dsize=1048576
\
vks4.home:/Copies/VMs \
/Users/Shared/NFS/VMs
My sequential read from NFS is as before (~50-70MiB/s). Which is "okay" for
that HW. But what can I do to bring the write speed back to 50MiB/s? Did I do
something wrong?
Thanks
--
You are receiving this mail because:
You are the assignee for the bug.