Hi

Ryusuke Konishi wrote:
> Hi David,
>
> On Wed, 17 Dec 2008 13:02:59 +0100 (CET), David Arendt wrote:
>   
>> Hi,
>>
>> I am using samba to serve files from a nilfs2 filesystem. From time to
>> time, I have smbd's hanging in state R with 100% cpu usage. Even kill
>> -KILL won't kill them. If I do an smbstatus it show me open files on the
>> nilfs2 filesystem which are not opened by any share. It started appearing
>> when I started using nilfs2, so I think it could be nilfs2 related but I'm
>> not sure. Did anyone experience similiar problems with nilfs2 and samba ?
>>     
>
> Let me ask you a few questions:
>
> First, please tell me your kernel version (and config if possible).
> Are you using nilfs-2.0.5 and nilfs-utils-2.0.6 packages?
>   
The kernel version is 2.6.27.9. I have attached the config file. I am
using the latest nilfs and nilfs-utils from git.
> How many snapshots (not checkpoints) did you create in the volume?
>   
The concerned volume is the only one where I don't do snapshots. I am
using this volume for video editing, so very large files are written to
it and many random reads are executed to different parts of these files.
The volume is 1tb in size.
> Does the cleaner (i.e. nilfs_cleanerd) affect on the problem?
> If you mount the volume with -i option, the cleanerd will not be
> invoked:
>
>  e.g.
>  # mount -t nilfs2 -i /dev/xxx /mount-dir
>
> This is convenient to separate cleaner problems.
>   
I did already test this and this didn't seem to make a difference.
However I'm wondering if it doesn't happen if only a few gbytes are left
on the drive as since 3 days I have more space on the concerned disk,
about 50 gbytes and the problem didn't occur until now.
> We are also using samba and nilfs in our storage server, so we have
> interest in the problem though we haven't yet experienced this
> phenomenon.
>
>
> With regards,
> Ryusuke
> _______________________________________________
> users mailing list
> users-JrjvKiOkagjYtjvyW6yDsg@public.gmane.org
> https://www.nilfs.org/mailman/listinfo/users
>   

Thanks in advance
Bye,
David Arendt