[mythtv-users] ramfs for mysql

[Simon Hobson]

Raymond Wagner wrote:

>Then you have bad information, that shouldn't exist at all, or should
>exist as a side note as "this is wrong" or "never do this".  Running
>MySQL in a tmpfs is of the last type.  It has been documented on the
>wiki several times in the past, and it has been rolled back or the page
>deleted each time.

Then I'd suggest the people deleting the page are idiots, and the 
sort of arrogant/aggressive people that give free software/open 
source projects such a bad reputation in some circles.

The fact that a page has appeared several times means that there 
**ARE** people who are thinking of this. As others have said, isn't 
it better to leave it there and put a BIG disclaimer at the top 
saying why it's bad ? At least then, people could find it, see the 
disclaimer and either : know that it's bad and forget about it, or 
make their own mind up as to whether it is a suitable option for them.
Instead, you seem to prefer hiding such information so people will 
guess in the dark, potentially oblivious to the risks and ways to 
mitigate them.


>But I only have the one hard drive.  ----  So go buy another to put the
>OS and database on, or scrounge up an old spare drive from somewhere.
>
>But I don't have slots on my board, or I don't have room in my case. 
>----  So buy a better machine to run your backend on.  That's why we
>recommend running the master backend on a real PC rather than a tiny
>embedded machine.
>
>I won't do it, I refuse.  ----  So turn off barriers, and continue
>operation as people did for years before switching to ext4 where
>barriers were enabled by default.  Live with the tiny risk that if you
>happen to be committing a transaction to disk the very instant the
>machine crashes, you end up with potential filesystem damage, and a
>database that needs repair.
>
>There is no situation where stuffing a MySQL server into volatile
>storage is the best solution.

Bollocks to that last bit. In effect you are saying that "if you 
won't buy 'approved' hardware, then p**s off" - that is the inference 
someone could draw. Refuse to support such setups, fine; but I really 
don't think it's your place to proscribe how people may (try to) use 
the software.

There are many many reasons why someone may wish to try things out on 
suboptimal hardware - I myself ran Myth backend for several years as 
a Xen guest on an underpowered machine, but I did that knowing the 
limitations and being prepared to live with them (until such time as 
I could afford/justify the additional hardware). I now have a 
dedicated backend, but had I not run Myth for a while on suboptimal 
hardware, then I'd probably have gone a different route altogether.

I'd say that running the DB in ram is no worse provided the person 
doing it understands the limitations. You seem to think that it's 
better to run a setup that can **CORRUPT** your database than to run 
a setup where the worst case is you lose transactions since your last 
backup point. Hmm, choose between unknown and potentially incidious 
corruption, or having your DB rolled back to a known good situation ?


So in summary, I think your position, and attitude, is just plain wrong.
-- 
Simon Hobson

Visit http://www.magpiesnestpublishing.co.uk/ for books by acclaimed
author Gladys Hobson. Novels - poetry - short stories - ideal as
Christmas stocking fillers. Some available as e-books.