Linux NFS development
 help / color / mirror / Atom feed
* file open for write fails under stress
@ 2008-10-13 22:51 Omry Yadan
       [not found] ` <48F3D106.9020203-FZzTFxfWWV5eoWH0uzbU5w@public.gmane.org>
  0 siblings, 1 reply; 7+ messages in thread
From: Omry Yadan @ 2008-10-13 22:51 UTC (permalink / raw)
  To: linux-nfs

I have an application that open many files for writing very fast, in a
way - it can almost be looked as a file system stress test.


my application is single threaded, and I noticed the following problem
when running a single instance of it:


at some point, after opening (and closing) many files for writing, the
application fails to open a file for writing.

if I sleep for 1 second and try again, it works.


my NFS server runs Debian etch 32 bit, with

nfs-kernel-server and nfs-common (1:1.0.10-6+etch.1).


client is also etch 32.

kernel on both is :2.6.18-6-686 #1 SMP Tue Jun 17 21:31:27 UTC 2008 i686


both machines are quad core.



I tried to mount with tcp and udp, and tried both sync and async export
point.


is this a known issue?

any known workaround that will not require messing up the application
with retry logic?


    Omry.


^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: file open for write fails under stress
       [not found] ` <48F3D106.9020203-FZzTFxfWWV5eoWH0uzbU5w@public.gmane.org>
@ 2008-10-14 13:23   ` Trond Myklebust
  2008-10-14 13:49     ` Omry Yadan
  2008-10-14 17:10   ` J. Bruce Fields
  1 sibling, 1 reply; 7+ messages in thread
From: Trond Myklebust @ 2008-10-14 13:23 UTC (permalink / raw)
  To: Omry Yadan; +Cc: linux-nfs

On Tue, 2008-10-14 at 00:51 +0200, Omry Yadan wrote:
> I have an application that open many files for writing very fast, in a
> way - it can almost be looked as a file system stress test.
> 
> 
> my application is single threaded, and I noticed the following problem
> when running a single instance of it:
> 
> 
> at some point, after opening (and closing) many files for writing, the
> application fails to open a file for writing.
> 
> if I sleep for 1 second and try again, it works.

What version of nfs, filesystem on the server, export options, mount
options,...?

Cheers
  Trond


^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: file open for write fails under stress
  2008-10-14 13:23   ` Trond Myklebust
@ 2008-10-14 13:49     ` Omry Yadan
       [not found]       ` <48F4A381.4040700-FZzTFxfWWV5eoWH0uzbU5w@public.gmane.org>
  0 siblings, 1 reply; 7+ messages in thread
From: Omry Yadan @ 2008-10-14 13:49 UTC (permalink / raw)
  To: Trond Myklebust; +Cc: linux-nfs

Trond Myklebust wrote:

> On Tue, 2008-10-14 at 00:51 +0200, Omry Yadan wrote:
>   
>> I have an application that open many files for writing very fast, in a
>> way - it can almost be looked as a file system stress test.
>>
>>
>> my application is single threaded, and I noticed the following problem
>> when running a single instance of it:
>>
>>
>> at some point, after opening (and closing) many files for writing, the
>> application fails to open a file for writing.
>>
>> if I sleep for 1 second and try again, it works.
>>     
>
> What version of nfs, filesystem on the server, export options, mount
> options,...?
>
>   

/etc/exports:
/var/vizi 38.99.126.160/28(rw,sync,no_subtree_check)
server file system : ext3 on a raid5 (900gb, 3x500gb hd).
server nfs version: 1.0.10

client mount:
nas1:/var/vizi on /var/vizi type nfs (rw,tcp,addr=38.99.126.163)


^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: file open for write fails under stress
       [not found]       ` <48F4A381.4040700-FZzTFxfWWV5eoWH0uzbU5w@public.gmane.org>
@ 2008-10-14 14:10         ` Trond Myklebust
  2008-10-14 14:46           ` Omry Yadan
  0 siblings, 1 reply; 7+ messages in thread
From: Trond Myklebust @ 2008-10-14 14:10 UTC (permalink / raw)
  To: Omry Yadan; +Cc: linux-nfs

On Tue, 2008-10-14 at 15:49 +0200, Omry Yadan wrote:
> Trond Myklebust wrote:
> 
> > On Tue, 2008-10-14 at 00:51 +0200, Omry Yadan wrote:
> >   
> >> I have an application that open many files for writing very fast, in a
> >> way - it can almost be looked as a file system stress test.
> >>
> >>
> >> my application is single threaded, and I noticed the following problem
> >> when running a single instance of it:
> >>
> >>
> >> at some point, after opening (and closing) many files for writing, the
> >> application fails to open a file for writing.
> >>
> >> if I sleep for 1 second and try again, it works.
> >>     
> >
> > What version of nfs, filesystem on the server, export options, mount
> > options,...?
> >
> >   
> 
> /etc/exports:
> /var/vizi 38.99.126.160/28(rw,sync,no_subtree_check)
> server file system : ext3 on a raid5 (900gb, 3x500gb hd).
> server nfs version: 1.0.10

> client mount:
> nas1:/var/vizi on /var/vizi type nfs (rw,tcp,addr=38.99.126.163)

So I assume it is NFS version 3, then?

What error does the application get when the open() fails?

Cheers
  Trond


^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: file open for write fails under stress
  2008-10-14 14:10         ` Trond Myklebust
@ 2008-10-14 14:46           ` Omry Yadan
  0 siblings, 0 replies; 7+ messages in thread
From: Omry Yadan @ 2008-10-14 14:46 UTC (permalink / raw)
  To: Trond Myklebust; +Cc: linux-nfs


>> /etc/exports:
>> /var/vizi 38.99.126.160/28(rw,sync,no_subtree_check)
>> server file system : ext3 on a raid5 (900gb, 3x500gb hd).
>> server nfs version: 1.0.10
>>     
>
>   
>> client mount:
>> nas1:/var/vizi on /var/vizi type nfs (rw,tcp,addr=38.99.126.163)
>>     
>
> So I assume it is NFS version 3, then?
>
> What error does the application get when the open() fails?
>   
according to nfsstat, yes - it's NFS v3.
after open fails (I use fstream binary_file(name,ios::out|ios::binary) )
the error message returned from strerror_r(errno, buf, 1000) is
something like "Input Output error".

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: file open for write fails under stress
       [not found] ` <48F3D106.9020203-FZzTFxfWWV5eoWH0uzbU5w@public.gmane.org>
  2008-10-14 13:23   ` Trond Myklebust
@ 2008-10-14 17:10   ` J. Bruce Fields
  2008-10-14 17:59     ` Omry Yadan
  1 sibling, 1 reply; 7+ messages in thread
From: J. Bruce Fields @ 2008-10-14 17:10 UTC (permalink / raw)
  To: Omry Yadan; +Cc: linux-nfs

On Tue, Oct 14, 2008 at 12:51:50AM +0200, Omry Yadan wrote:
> I have an application that open many files for writing very fast, in a
> way - it can almost be looked as a file system stress test.
> 
> 
> my application is single threaded, and I noticed the following problem
> when running a single instance of it:
> 
> 
> at some point, after opening (and closing) many files for writing, the
> application fails to open a file for writing.

Do you know what error exactly the open is failing with?

--b.

> if I sleep for 1 second and try again, it works.
> 
> 
> my NFS server runs Debian etch 32 bit, with
> 
> nfs-kernel-server and nfs-common (1:1.0.10-6+etch.1).
> 
> 
> client is also etch 32.
> 
> kernel on both is :2.6.18-6-686 #1 SMP Tue Jun 17 21:31:27 UTC 2008 i686
> 
> 
> both machines are quad core.
> 
> 
> 
> I tried to mount with tcp and udp, and tried both sync and async export
> point.
> 
> 
> is this a known issue?
> 
> any known workaround that will not require messing up the application
> with retry logic?
> 
> 
>     Omry.
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: file open for write fails under stress
  2008-10-14 17:10   ` J. Bruce Fields
@ 2008-10-14 17:59     ` Omry Yadan
  0 siblings, 0 replies; 7+ messages in thread
From: Omry Yadan @ 2008-10-14 17:59 UTC (permalink / raw)
  To: J. Bruce Fields; +Cc: linux-nfs

J. Bruce Fields wrote:

> On Tue, Oct 14, 2008 at 12:51:50AM +0200, Omry Yadan wrote:
>   
>> I have an application that open many files for writing very fast, in a
>> way - it can almost be looked as a file system stress test.
>>
>>
>> my application is single threaded, and I noticed the following problem
>> when running a single instance of it:
>>
>>
>> at some point, after opening (and closing) many files for writing, the
>> application fails to open a file for writing.
>>     
>
> Do you know what error exactly the open is failing with?
>   
As I wrote in a separate message to the list:
the error string is "Input Output error".
I can get the exact errno if you like, but it's probably a generic IO
error code.

^ permalink raw reply	[flat|nested] 7+ messages in thread

end of thread, other threads:[~2008-10-14 18:00 UTC | newest]

Thread overview: 7+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2008-10-13 22:51 file open for write fails under stress Omry Yadan
     [not found] ` <48F3D106.9020203-FZzTFxfWWV5eoWH0uzbU5w@public.gmane.org>
2008-10-14 13:23   ` Trond Myklebust
2008-10-14 13:49     ` Omry Yadan
     [not found]       ` <48F4A381.4040700-FZzTFxfWWV5eoWH0uzbU5w@public.gmane.org>
2008-10-14 14:10         ` Trond Myklebust
2008-10-14 14:46           ` Omry Yadan
2008-10-14 17:10   ` J. Bruce Fields
2008-10-14 17:59     ` Omry Yadan

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox