NetBSD-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Very poor NFS I/O



Hauke Fath a écrit :
> On 2/27/26 14:27, BERTRAND Joël wrote:
>> - /, /usr, /usr/src, /var and /srv on raid0 (Raid-1 volume) ;
>> - /home on raid1 (Raid5 volume) ;
>> - some others tmp filesystems on a ccd0 device exported through iSCSI.
>>
>>     All filesystems are FFSv2+EA.
>>
>>     CPU is an i7-4770, main memory 16 GB. This server exports /srv and
>> /home through NFS (V3/TCP, 128 threads, async) and disk I/O from NFS
>> clients are very slow. Server load can raise until 110 or 120 during
>> huge NFS access.
> 
> How many clients?

	4/5 (FreeBSD, Linux and OpenVMS), all clients are diskless workstations.

> Are the server disks spinning rust, or solid state?

	Regular disks, not solid state.

	For example : /home is dk0.

legendre# dkctl dk0
strategy:
/dev/rdk0: readprio

getcache:
/dev/rdk0: read cache enabled
/dev/rdk0: write-back cache enabled
/dev/rdk0: read cache enable is not changeable
/dev/rdk0: write cache enable is changeable
/dev/rdk0: cache parameters are not savable
/dev/rdk0: cache Force Unit Access (FUA) supported
/dev/rdk0: cache Disable Page Out (DPO) not supported

listwedges:
dkctl: /dev/rdk0: listwedges: Inappropriate ioctl for device
legendre#  raidctl -s raid1
Components:
           /dev/wd4e: optimal
           /dev/wd6e: optimal
           /dev/wd5e: optimal
...
legendre# dkctl wd3
strategy:
/dev/rwd3d: priocscan
...
legendre# dkctl wd4
strategy:
/dev/rw4d: priocscan
...
legendre# dkctl wd5
strategy:
/dev/rwd5d: priocscan
...

> I have been running an nfs server for networked /home for more than 20
> years, and have been avoiding raid5 (in favour of mirrored disks) for
> almost as long.
> 
> You want as many transactions/sec as you can get.

Attachment: signature.asc
Description: OpenPGP digital signature



Home | Main Index | Thread Index | Old Index