Hi!
>>>>> "Joerg" == Joerg Bruehe <Joerg.Bruehe@stripped> writes:
<cut>
Joerg> I have now finished and checked a fix that will make the installation
Joerg> and upgrade of the MySQL server (via RPM packages) behave as follows:
Joerg> - On first installation, the start/stop scripts will be installed in
Joerg> "/etc/init.d" and symlinked to the run level directories,
Joerg> which will cause the MySQL server to be started whenever the machine
Joerg> boots up.
Joerg> However, to do as demanded in the bug report, the MySQL server will
Joerg> not be started directly during installation, but at the next reboot
Joerg> (unless the sysadmin changed the symlink entries to the run level
Joerg> directories).
Joerg> - On an upgrade, the "pre" section of the RPM spec file will check
Joerg> whether the MySQL server is running or not.
Joerg> It will, using an entry in a file, transfer this knowledge to the
Joerg> "post" section of the spec file, and then stop the server.
Joerg> Then, the old binaries will be replaced by the new ones, as usual.
Joerg> The "post" section will check that file entry:
Joerg> If the server process was running before the upgrade, the new binaries
Joerg> will be autostarted during the upgrade step.
Joerg> If the server process had been stopped before the upgrade, no
Joerg> autostart will be done, but the start/stop script will be entered into
Joerg> the system - so as with installation, the server will be started on
Joerg> the next reboot.
Joerg> In short words: After an upgrade (via RPM), the MySQL server will be
Joerg> running if and only if it was running before the upgrade.
Joerg> My questions to all interested observers, both inside and outside the
Joerg> company:
Joerg> - Do you agree to this behaviour, or do you see any problems with it?
Looks very good.
Joerg> - I would like to introduce the changed upgrade policy with the next
Joerg> release builds in 5.1 and 5.5 after this change got approved,
Joerg> but introduce the changed installation policy in 5.5 only.
Joerg> My reasoning is that not autostarting on installation is a functional
Joerg> change (however desirable), but not autostarting a stopped server (on
Joerg> upgrade) is a bug fix.
Joerg> If there are any objections to this, I would like to hear of them,
Joerg> rather than disappoint you later.
Ok with me.
Regards,
Monty
