[mythtv-users] Repair mysql tables after power failure?

[R. G. Newbury]

Mike Perkins wrote:
> David Brodbeck wrote:
>> On Oct 3, 2007, at 6:51 AM, Mike Perkins wrote:
>>> Hmmm. Note to self: add this to startup on the master backend, in  
>>> such a way it
>>> runs before the backend starts. The master backend machine  
>>> currently stays up
>>> forever, unless there's a power outage. Although I run a script  
>>> containing this
>>> around 02:00 every night, I hadn't considered possible corruption  
>>> after a power
>>> cut/restart. Filesystems get checked on startup, why not databases?
>>>
>>> In fact, does Mysql do anything useful on startup in this regard?
>> I don't think it does by default.  Some distributions include a table  
>> check in their startup script.  I like to add one, for the reasons  
>> you state.
>>
>> If you want to get really fancy, create a file in the startup script  
>> and delete it in the shutdown script, then only do the check on  
>> startup if the file is present.  This would mimic the way fsck will  
>> only check "dirty" filesystems.
>>
> Consider it done. Is it worth adding something of this kind to Mythtv startup 
> generally? OK, I get a couple of outages/year so it's not a big problem 
> (actually, night before last, had 5 in 15 minutes, which is why this is in the 
> front of my brain just now). But power outages aren't the only reason a backend 
> can fall over. Should we be considering a database check every time the backend 
> comes up, if the above-mentioned "dirty" flag exists? i.e. built in to the 
> backend startup and shutdown scripts?
> 
> Mike Perkins

Yes, please! I was thinking about this over lunch and it occurs to me 
that a standalone optimize/check script which could be run be the 
start/stop service scripts would also work as a cron script, with a 
little care. Among other things, it should note the presence of the 
'dirty-bit' flag and restore that state, if run from a cron.

Hmm, on second thought, it should ignore the present state, and *set* 
the flag, immediately on entry, whether as startup or as a cron job, in 
case it fails, or the power cycles, while it is running.

Then the start script needs only a one-line call to the script, while 
shutdown is a one-line 'rm -f...'


              R. Geoffrey Newbury			

Tux says: "Be regular. Eat cron flakes."