paul-aviles
commented
9 years ago That is probably the most important one. Default value is 8. You can see the number of threads using:
ps aux | grep nfs.
root 1554 0.0 0.0 0 0 ? S< Jul20 0:00 [nfsd4] root 1555 0.0 0.0 0 0 ? S< Jul20 0:00 [nfsd4_callbacks] root 1559 0.0 0.0 0 0 ? S Jul20 40:15 [nfsd] root 1560 0.0 0.0 0 0 ? S Jul20 40:11 [nfsd] root 1561 0.0 0.0 0 0 ? S Jul20 39:54 [nfsd] root 1562 0.0 0.0 0 0 ? S Jul20 40:19 [nfsd] root 1563 0.0 0.0 0 0 ? S Jul20 40:28 [nfsd] root 1564 0.0 0.0 0 0 ? S Jul20 40:08 [nfsd] root 1565 0.0 0.0 0 0 ? S Jul20 40:20 [nfsd] root 1566 0.0 0.0 0 0 ? S Jul20 40:29 [nfsd] root 18615 0.0 0.0 112644 968 pts/1 S+ 12:40 0:00 grep --color=auto nfs
The number of [nfsd] entries will be the one to note, 8 in this case.
As fas as performance is concerned other things to consider is the mount options on /etc/fstab.
On my NFS server which provides VMware backend storage I use the following on /etc/fstab for the mount partition that holds the NFS data:
btrfs subvol=nas,nodatacow,noatime,nodiratime,autodefrag 1 2
The /etc/exports I have the following: /nas/200/ARCHIVE 172.16.1.0/24(rw,async,no_root_squash,no_all_squash)
This way the system does not need to update time stamps for irrelevant stuff, but that may be desired if using NFS as a regular filer.
Paul Aviles
support some parameters for the service. One suggested is RPCNFSDCOUNT