[mythtv-users] Storage groups, NFS, and uniqueness

[Jim Stichnoth]

On Wed, Aug 19, 2009 at 8:17 PM, Michael T. Dean<mtdean at thirdcontact.com> wrote:
> On 08/19/2009 09:36 PM, Jim Stichnoth wrote:
>>
>> On Wed, Aug 19, 2009 at 5:41 PM, Michael T. Dean wrote:
>>
>>>
>>> On 08/19/2009 08:18 PM, Michael T. Dean wrote:
>>>
>>>>
>>>> On 08/19/2009 08:04 PM, Jim Stichnoth wrote:
>>>>
>>>>>
>>>>> I have a master backend (mythmaster) with two drives, and the default
>>>>> storage group has two directories, one per drive.  I have a slave
>>>>> backend (mythfe3), which has a USB tuner and is diskless.  Its default
>>>>> storage group is set up to point over NFS to the master backend's two
>>>>> drives.
>>>>>
>>>>> Is there some way to let the system know that the storage groups on
>>>>> the master and slave backend are really the same, for the purpose of
>>>>> disk load balancing and free space calculation?  For example, the
>>>>> mythweb backend status page shows twice the amount of available space
>>>>> as there actually is:
>>>>>
>>>>
>>>> http://www.gossamer-threads.com/lists/mythtv/dev/392375#392375 and the
>>>> rest of the thread.  It's fixed in trunk, now.
>
> ...
>>>
>>> Can you please post the output of:
>
> ...
>>
>> Thanks.  I will try the "unoptimized" NFS.  Here is the df output, run
>> at the same time while no recordings were in progress:
>
> Thanks.  That was exactly what I needed.  The fix went into 0.21-fixes in
> r21368 , so once that version is available from your packager and you've
> updated, you can re-optimize your NFS (take out the 32kiB blocksize).
>
>> Another issue is that thanks to autofs, I often get reports like this:
>>
>> Disk Usage:
>> ...
>>    * MythTV Drive #3:
>>          o Directories: mythfe3:/net/mythmaster/storage/recordings,
>> mythfe3:/net/mythmaster/storage1/recordings
>>          o Total Space: -1 MB
>>          o Space Used: -1 MB
>>          o Space Free: 0 MB
>> ...
>>
>> It probably wouldn't hurt to probe the directory before doing the "df"
>> equivalent to be sure the directory is auto-mounted.
>
> OK, so in the interest of full disclosure, I know nothing about autofs, so
> I'm just the messenger, here, but we have some code in Myth (both 0.21-fixes
> and trunk) that deals with autofs and works for the Chris Pinkham, the
> developer who wrote it.  He mentioned, though, that the issue you're seeing
> seems to be caused by using --ghost with autofs.

Thanks again.  Switching to a 32KB block size works like a charm.  I
don't know anything about a "--ghost" option, but I added a
"--timeout=0" option to prevent it from unmounting automatically.

Jim