Stuart Henderson wrote:

quoted text
>

> On 2007/09/19 16:28, Tomas wrote:

> >   That's an excelent idea. I will use 'pkill mysqld'. But instead

> > of 'sleep 10' myabe it's better to check every second if mysqld.sock

> > exists?

>

> I would choose to limit the maximum time, in case mysqld deadlocks.

>

> I think "pgrep mysqld" is a safer test than mysqld.sock (unless

> you've read mysql code and determined that the control socket is

> only removed after the tables have been flushed).

>

Something you should be aware of:

If there is insufficient space, MySQL will wait patiently for the

space to become available at which point it picks up like there

never was a problem.

Current behavior may be different. YMMV etc.

Surprising, but it can allow manipulation in low-disk situations.

I've no idea what happens to pending transactions (non-MyISAM databases)