NFS Optimization: Difference between revisions
Jump to navigation
Jump to search
No edit summary |
No edit summary |
||
| (One intermediate revision by the same user not shown) | |||
| Line 1: | Line 1: | ||
===Server Side=== | ===Server Side=== | ||
====/etc/exports==== | ====/etc/exports==== | ||
< | <syntaxhighlight lang=bash> | ||
/mnt/raid5 192.168.15.142/32(rw,async,no_root_squash) | |||
</syntaxhighlight> | |||
async - dramatic throughput increase, but dangerous if a client does not unmount cleanly.. | async - dramatic throughput increase, but dangerous if a client does not unmount cleanly.. | ||
====Tuning /etc/sysctl.conf==== | ====Tuning /etc/sysctl.conf==== | ||
< | <syntaxhighlight lang=bash> | ||
net.core.rmem_default = 262144 | net.core.rmem_default = 262144 | ||
net.core.rmem_max = 262144 | net.core.rmem_max = 262144 | ||
| Line 12: | Line 14: | ||
net.ipv4.ipfrag_high_thresh = 524288 | net.ipv4.ipfrag_high_thresh = 524288 | ||
net.ipv4.ipfrag_low_thresh = 393216 | net.ipv4.ipfrag_low_thresh = 393216 | ||
</ | </syntaxhighlight> | ||
< | <syntaxhighlight lang=bash> | ||
echo 0 > /proc/sys/net/ipv4/tcp_sack | echo 0 > /proc/sys/net/ipv4/tcp_sack | ||
echo 0 > /proc/sys/net/ipv4/tcp_timestamps | echo 0 > /proc/sys/net/ipv4/tcp_timestamps | ||
</ | </syntaxhighlight> | ||
====TCP Segmentation offload==== | ====TCP Segmentation offload==== | ||
This will take off some of the tcp overhead if your card supports it.. | This will take off some of the tcp overhead if your card supports it.. | ||
< | <syntaxhighlight lang=bash> | ||
# ethtool -K ethN tso on | |||
</syntaxhighlight> | |||
===Client Side=== | ===Client Side=== | ||
=====/etc/fstab===== | =====/etc/fstab===== | ||
< | NFSv4 Client: | ||
[rw]size=32768 - NFSv3 maximum read write size | <syntaxhighlight lang=bash> | ||
intr - if the mount drops, you'll still be able to ^C out of whatever operation your running | 192.168.15.20:/mnt/raid5 /mnt/raid5 nfs defaults 0 0 | ||
hard - | </syntaxhighlight> | ||
NFSv3 Client: | |||
<syntaxhighlight lang=bash> | |||
192.168.15.20:/mnt/raid5 /mnt/raid5 nfs rsize=32768,wsize=32768,intr,hard 0 0 | |||
</syntaxhighlight> | |||
[rw]size=32768 - NFSv3 maximum read write size<br/> | |||
intr - if the mount drops, you'll still be able to ^C out of whatever operation your running<br/> | |||
hard - flush file locks before being able to unmount<br/> | |||
Latest revision as of 01:33, 25 January 2018
Server Side
/etc/exports
/mnt/raid5 192.168.15.142/32(rw,async,no_root_squash)
async - dramatic throughput increase, but dangerous if a client does not unmount cleanly..
Tuning /etc/sysctl.conf
net.core.rmem_default = 262144
net.core.rmem_max = 262144
#
# Increase the fragmented packet queue length
net.ipv4.ipfrag_high_thresh = 524288
net.ipv4.ipfrag_low_thresh = 393216
echo 0 > /proc/sys/net/ipv4/tcp_sack
echo 0 > /proc/sys/net/ipv4/tcp_timestamps
TCP Segmentation offload
This will take off some of the tcp overhead if your card supports it..
# ethtool -K ethN tso on
Client Side
/etc/fstab
NFSv4 Client:
192.168.15.20:/mnt/raid5 /mnt/raid5 nfs defaults 0 0
NFSv3 Client:
192.168.15.20:/mnt/raid5 /mnt/raid5 nfs rsize=32768,wsize=32768,intr,hard 0 0
[rw]size=32768 - NFSv3 maximum read write size
intr - if the mount drops, you'll still be able to ^C out of whatever operation your running
hard - flush file locks before being able to unmount