>>>>> "Tim" == Tim Bunce <Tim.Bunce@stripped> writes:
Tim> On Thu, Jan 06, 2000 at 10:48:47PM +0200, Michael Widenius wrote:
>> >>>>> "Tim" == Tim Bunce <Tim.Bunce@stripped> writes:
Tim> Also, something I'm not sure of... does MySQL mark table (and index)
Tim> files as being 'stable' when they are closed and check that flag when
Tim> they are reopened? (Like most modern filesystems do?)
>> The new MyISAM does the first part; As soon as we have myisamchk
>> integrated into mysqld we will handle the second part (automatic
>> checking + repair of tables at reboot)
Tim> But meanwhile shouldn't MySQL fail to open (or at least change) a table
Tim> that's potentially corrupt?
As an integrated myisamchk is only a couple of releases away we
haven't done that. You can instead quickly detect and repair not
properly closed MyISAM files with:
myisamchk --force --fast --quick *.MYI
The above will ignore all tables that tables that hasn't been modified
since last check and it will at once notice tables that arn't closed properly.
Tim> If so, if that flag set by just flushing a table?
>> Yes, the flag is set by doing a flush on the table (as the flush
>> forces a close of all tables + a reopen on the locked tables).
Tim> So a LOCK + FLUSH + filecopy is a reliable way to make a 'stable'
Tim> copy of a table.