Prev: "Hadron" is a little lying nebbish
Next: mdadm (SW-RAID) and RAID6 performance. Or should I use HW-RAID?
From: J G Miller on 6 Aug 2010 09:36
On Fri, 06 Aug 2010 07:22:31 +0000, Huge wrote:
>
> Hard NFS mounts are not necessarily a good idea. If the server hangs, so
> will the client.

Exactly. If you are doing read-only mounting, then mounting with the hard
option is less necessary, and if using NFS v4 with TCP even less critical
again.

From man 5 nfs

NB: A so-called "soft" timeout can cause silent data
corruption in certain cases. As such, use the soft
option only when client responsiveness is more important
than data integrity. Using NFS over TCP or increasing
the value of the retrans option may mitigate some of the
risks of using the soft option.

If rriting data to an NFS mounted disk, then it is much safer to use
hard to avoid possible data corruption issues, but if your data is
really that critical, you should not be writing it to an NFS mounted disk,
eg save locally and then use rsync, and obviously for multi client
operation you would be using a database anyways ...
From: Rahul on 6 Aug 2010 17:20
Huge <Huge(a)nowhere.much.invalid> wrote in news:8c1rhnFc72U1
@mid.individual.net:

> Hard NFS mounts are not necessarily a good idea. If the server hangs, so
> will the client.
>
>

True. But doesn't the seciton of the NFS manuals also reccomend
hard mounts to prevent data loss?

>As such, use the soft option only when client responsiveness is more
>>important than data integrity.

So, does one have to chose between the devil (data loss) and the deep sea
(hung client)?

I thought the way to go was using hard + intr to get both integrity and
responsiveness. Unfortunately never got it to work that way but that's
another story.

--
Rahul
From: zedkay on 9 Aug 2010 08:34
On 08/06/2010 03:36 PM, J G Miller wrote:
> On Fri, 06 Aug 2010 07:22:31 +0000, Huge wrote:
> If writing data to an NFS mounted disk, then it is much safer to use
> hard to avoid possible data corruption issues, but if your data is
> really that critical, you should not be writing it to an NFS mounted disk,
> eg save locally and then use rsync, and obviously for multi client
> operation you would be using a database anyways ...

Actually, one should a clustered filesystem,
http://en.wikipedia.org/wiki/Clustered_file_system
and then preferably in a SAN environment.

--
Please do not reply to my Email address. It is a faux Email address.
Cyberpunk FPS/MMORG www.neocron.com
Runs on Windows, platinum in latest WINE/Ubuntu. Running since 2002.
From: Ignoramus29207 on 9 Aug 2010 11:35
On 2010-08-09, zedkay <zedkay(a)maileater.com> wrote:
> On 08/06/2010 03:36 PM, J G Miller wrote:
>> On Fri, 06 Aug 2010 07:22:31 +0000, Huge wrote:
>> If writing data to an NFS mounted disk, then it is much safer to use
>> hard to avoid possible data corruption issues, but if your data is
>> really that critical, you should not be writing it to an NFS mounted disk,
>> eg save locally and then use rsync, and obviously for multi client
>> operation you would be using a database anyways ...
>
> Actually, one should a clustered filesystem,
> http://en.wikipedia.org/wiki/Clustered_file_system
> and then preferably in a SAN environment.
>

Guys, with all due respect for these ideas, I would really to find an
answer to my problem. The key to it seems to be this message in dmesg:

[44743.592195] nfs: server myserver not responding, still trying
[45103.592528] nfs: server myserver OK
[45980.844190] nfs: server myserver not responding, still trying

The message is not really true, the servers (at least at the time of
checking) was accessible.

i
From: Ignoramus29207 on 9 Aug 2010 12:01
On 2010-08-05, Rahul <nospam(a)nospam.invalid> wrote:
> Ignoramus16841 <ignoramus16841(a)NOSPAM.16841.invalid> wrote in news:5
> _KdnW29vrfPisbRnZ2dnUVZ_rmdnZ2d(a)giganews.com:
>
>> Could be some sort of timing bug in the NFS server code?
>>>
>>
>> It could be anything, but I am more inclined to blame the client, as
>>
>> 1) Many clients work just fine
>> 2) Even my bad desktop works fine when it connects after a reboot
>>
>
> Does it matter if it is a soft or hard mount?
>
> I'm just guessing here. I've a long mount option string on my server:
>
> eustorage:/opt /opt nfs rw,nodev,noatime,nfsvers=3,timeo=110,retrans=
> 50,hard,intr,proto=udp,rsize=32768,wsize=32768 0 0
>
> But I'd be hard pressed to explain why I chose each option that I have in
> there. :)
>
>

None of this really makes any difference.

i
First  |  Prev  |  Next  |  Last
Pages: 1 2 3 4 5
Prev: "Hadron" is a little lying nebbish
Next: mdadm (SW-RAID) and RAID6 performance. Or should I use HW-RAID?