[mythtv-users] Slave backend with shared NFS storage thinks it has its own disk.

[Tim Phipps]

On Thursday 26 April 2007 8:14 pm, Chris Pinkham wrote:
> * On Thu Apr 26, 2007 at 07:13:15PM +0100, Tim Phipps wrote:
> > Filesystem           1K-blocks      Used Available Use% Mounted on
> > /dev/mapper/vg1-lv1  459100320 360828500  93607624  80% /home
> >
> > nfs:/home            459100320 360828480  93607648  80% /home
>
> I believe this is similar to what people were seeing.  It was fixed by
> changing the order of some multiplication and division in the code.  After
> I put the fix in -fixes earlier this week, Jarod merged it into the 0.20.1
> tag branch, so it will be in 0.20.1 when that version is released.
>
> > out by 20k/24k. Could this be due to different block sizes? If so how do
> > I set the blocksize on an NFS mount?
>
> I think nfs sets the block size to be one of the read/write sizes, but I'm
> not sure.  You can set these using "-o rsize=8192,wsize=8192" or something
> similar on your nfs mount command.
>
That fixed it. Setting rsize/wsize=4k made the df outputs match and the 
frontend status screen lump the two backends together. The rsize/wsize was 
32k so this might not fix my problems with streaming, I'll let you know after 
the weekend. There are other things I must try out, like noatime and 
journalling options.

Cheers,
Tim.