informations about boot sequence (Re: F15 - mysql start problem)

[Lennart Poettering]

On Mon, 09.05.11 18:58, Michał Piotrowski (mkkp4x4 at gmail.com) wrote:

> >> Yes, mysql and postgresql are both certainly broken by this.  Please
> >> send the info, I'll take care of those two.
> >
> > I'll create bug reports and send proposed patches later.
> >
> 
> Patch for MySQL
> https://bugzilla.redhat.com/show_bug.cgi?id=703214
> 
> for PostgreSQL
> https://bugzilla.redhat.com/show_bug.cgi?id=703215
> 

Urks. I would strongly suggest not to make changes like this by
default. These services should bind on 0.0.0.0 by default, which is
available without network. Only if you manually configure them to bind
on a specific IP address you need the network up and configured. But not
even that is strictly necessary. A nice server uses netlink to subscribe
to network changes and binds as appropriate. A simpler server could even
invoke the IP_FREEBIND setsockopt (1-line patch), which allows them to
listen on the addresses before the interfaces are up (which you probably
want to do anyway, since it adds robustness and makes your server
independent from network changes).

So, instead of making mysql/postgres start up slower for everybody, i'd
rather see this solution:

a) people who manually change the IP address fo mysql to bind on
specific ip addresses, manually also enable
NetworkManager-wait-online.service.

or:

b) the servers are fixed to listen to netlink.

or:

c) They get fixed to use IP_FREEBIND.

All three of the solutions are nicer than adding unnecessary
dependencies for them.

Lennart

-- 
Lennart Poettering - Red Hat, Inc.