Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 4 Apr 2013 23:48:19 +0200
From:      b w <bw.mail.lists@gmail.com>
To:        freebsd-questions@freebsd.org
Subject:   NFSv4 questions and possible bugs
Message-ID:  <CAHM0YgttPgg9gBqZLp18ix4ZpX7zv2AOA5870HP8770QKSYc2w@mail.gmail.com>

next in thread | raw e-mail | index | archive | help
I set up NFSv4, did some performance tests, setup looks like this:

Server rc.conf:
nfs_server_enable="YES"
nfsv4_server_enable="YES"
nfsuserd_enable="YES"

exports:
/share -mapall=nobody  10.10.14.2 10.10.14.3
V4: /   -sec=sys

Client(s) fstab mount:
srv:/share /mnt nfs nfsv4,rsize=32768,wsize=32768,rw   0   0

Server is in a different vlan than the clients, there's a Juniper SRX
between them.

As far as I understand this means a NFSv4 only setup.

1. I had to use rsize and wsize mount options, without them performance
is horrible, 1 MBps from the same vlan, when in different vlans it would
start fast than drop to a standstill, compared to around 100MBps with sizes. Not
sure why. 32K is the best I found, 16K and 64K were slightly worse, but
I assume this is due to our network setup.

2. Only port 2049 is open in the firewall, as it should be enough for
NFSv4, but umount tries to send 3 UDP packets to port 111. This causes
it to hang for some time while waiting for the packets to time out and
exit with an error. The unmount is executed correctly, but the exit
status could cause problems in scripts, see 4.

3. bonnie++ exits uncleanly,
http://lists.freebsd.org/pipermail/freebsd-current/2010-September/019820.html
I guess this is a known bug, but I just wanted to point out that it's
still there in up to date 9.1-RELEASE. Since it's been around for a
long time, I suppose it's not likely to cause problems in production,
is it?

4. After bonnie++'s failure I tried iozone, but iozone wants to
unmount before each test and hits #2.

Performance is excellent as far as I can see, after setting raise and
wsize, transfers hit the network cap, so I guess my main question is
if #2 is likely to cause issues down the road. It will have mostly
perl scripts reading and moving files around and syslog, "rm -rf"
seemed to do the job without problems.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAHM0YgttPgg9gBqZLp18ix4ZpX7zv2AOA5870HP8770QKSYc2w>