[mythtv-users] slave backend attached via wireless 802.11g
ajlill at ajlc.waterloo.on.ca
Tue Dec 8 17:10:54 UTC 2009
Glen Dragon <gdragon at jetcom.org> writes:
> For logistical reasons (and because I never know when to stop playing) I
> have moved and setup a slave backend attached to a hd-pvr, but the trick is
> that it is attached to the network only via 802.11g wireless.
> For the most part this works pretty well; actually better than I expected.
> Not all that surprisingly, the general operation of the slave backend works
> well.. the database accesses, recording commands, etc
> Streaming from that slave to other frontends is not completely terrible.. it
> works without loss sometimes, but other times it drops packets (likely due
> to other 2.4 Ghz interference)
If you transcoded, then you could eliminate this. I run a frontend on
my laptop, and I can only reliably stream if I transcode first.
> Currently I have it recording to a local (300 GB) hdd. I have no intention
> of storing the recordings there long term.. I would copy them to the master
> BE, w/ 6 TB of storage.
> My current thought is to have a post-recording job that will:
> - copy [rsync?] the completed recordings to the MBE
> - update the database to transfer ownership to the MBE.
> - upon successful transfer delete the recording from the slave's hdd
> - commflag ??? on which machine?
I have a job that looks for recording files, check the inuseprogram
table to see if something is using it , and if not, does the
copy. Probably better than rsync. Now I'm on a wired net, so I let my
master commflag and transcode over NFS, but this is pretty much what I
do on my setup. I can send you some of my scripts if you like.
> I'm currently doing commflagging on the slave, but it's not fast enough to
> do it real-time. I'm not sure if it's better to let the MBE do the
> commflagging after the file copy is done.
> Is it possible to specify that a User job would not run until the
> commflagging is complete? If I wait until the copy to the MBE is complete,
> how do i create a new job on the MBE to comm flag that recording?
I noticed that 0.22 is running all user jobs in parallel, even on the
same recording, so I'm not sure how, of if you can, serialize them,
unless you turn down the number of parallel jobs to 1. I know that in
0.21, you could disable jobs from running on certain machines, so you
can force all commflagging to happen on the backend. Now, the problem
is that you can't run a use job before commflag and transcode, AFAIK.
I suppose you could make userjob 1 do the transfer, and userjob 2
> I considered mounting the MBE storage via NFS to the slave, but I'm not sure
> that the reliability and performance of the netmount would not get in the
> way of the recordings. I prefer that the recordings complete successfully,
> and then transfer them non-realtime.
Definite no-no, at least over wireless. You could commflag that way
(mount the slave fs on the master), but it would probably still be too
slow for realtime.
> I'm wondering is anyone has tried this, or has any best practices..
> mythtv-users mailing list
> mythtv-users at mythtv.org
More information about the mythtv-users