[mythtv-users] how to get slaves to automatically reconnect to the master backend

[David Snider]

On Aug 20, 2009, at 10:12 AM, Michael T. Dean wrote:

> On 08/20/2009 10:28 AM, Peter Abplanalp wrote:
>> i ran a test and stopped the master backend.  i got a bunch of lines
>> like the following on one slave:
>>
>> 2009-08-20 08:18:23.830 Connecting to master server:  
>> 192.168.1.104:6543
>> 2009-08-20 08:18:23.846 Connection to master server timed out.
>>
>> once i started the backend again, i got a 'connected successfully'
>> message.  i also checked the status page on mythweb and it reported
>> that all the tuners on all the backends were available so it would
>> seem that upon graceful shutdown and startup the slaves to reconnect.
>> i also ran 'kill -9 <mythbackend pid>' to try and simulate a crash.
>> in that case, the slaves also reconnected to the master.
>>
>> i wonder if there is any way to get the slaves to reconnect after a
>> reboot or what the difference is in that case.
>
> My guess would be the difference is that when the master backend  
> crashes, it's not the only thing that crashes--i.e. you're probably  
> running the MySQL server on the master backend host, right?  When  
> the system goes down, it takes the master mythbackend and the MySQL  
> server down.  Myth does not deal well with dying MySQL.  Try killing  
> both the master backend and the database server.
>
> I'd recommend a) changing a remote backend to become the master  
> backend and changing the master backend to become a remote backend,  
> and b) moving the MySQL server to one of the remote backend hosts,  
> then c) taking however long it takes to figure out the problem  
> causing the master backend to reboot and in the process only worry  
> about ever losing one capture device.
>
This might be overkill, but I use Nagios to check the tuner status,  
then have it configured to restart the service if there is an issue.    
FWIW, my backends don't always reconnect, even on a graceful restart  
of the master.