hi david, caleb, On Thu, Jan 05, 2006 at 12:39:47AM -0500, Caleb Epstein wrote: > The manual fix is to run "mysql_convert_table_format mysql" to > upgrade your "mysql" database to the MyISAM table format and > then the upgrade to 5.0 should go smoothly. > > If you have any data at risk, run "mysqldump --opt > --all-databases" so you can recover your data if something > tragic happens. > > Maintainers: perhaps the preinst should check for any > databases having .IS[MD] files and abort if they are found, > warning the user to run mysql_convert_table_format or possibly > running it for them before stopping and purging the old 4.1 > server?
having spent a bit of time thinking about it, it's a touch more
complicated than it may seem at first glance. here are some
potential gotchas:
- the best/only reliable place to run this is in new.preinst.
- as a result, we can only use files/scripts that exist in the
older version.
- this also means we can not say "we will now fail and you should
read the contents of /usr/share/doc/foo" (though we could refer
them to rtfm since the tfm already exists)
- mysql_convert_table requires passing sensitive data on the cmdline
given all this i think the following should cover it:
[new.preinst]
- determine list of dbs/tables containing old format, ignore if empty
- debconf prompt to determine whether to abort/fix/ignore
- if fix
- start mysql (should be stopped by old.prerm)
- issue ALTER TABLE commands manually using debian-sys-maint
account
- if debian-sys-maint can not connect, inform the
admin and then fall back as if "abort" had been
chosen.
- stop mysql server
so then the only thing that's missing is the actual code, and
some debconf templates. christian: what do you think?
could one of you guys provide me with the output from find /var/lib/mysql,
to make my testing just a bit faster?
thanks
sean
--
signature.asc
Description: Digital signature
