First of all please note that i really appreciate the tomcat developers work
and thank them
for it.
However see my notes below:


On Mon, Oct 26, 2009 at 9:21 PM, Christopher Schultz <
ch...@christopherschultz.net> wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Tsirkin,
>
> On 10/25/2009 8:10 AM, Tsirkin Evgeny wrote:
>
> Instead, it's a simple
> case of choosing a single node to act as the controller for the cluster.
> Just add a configuration parameter to the <Cluster> element to identify
> one server as the controller and have /that/ one manage the sessions
> when the cluster is shut down.
>
> Hang on, there's no "shutdown cluster" message that all nodes
> understand. So, let's add that. But wait, what happens if the cluster
> controller goes down? We need a backup for the cluster controller. Let's
> add a backup for the cluster controller, and then a backup for the
> backup, and then a backup for the backup for the backup...
>
>

 Actually your solution is pretty fine. Just that there is no need in
"controller" backup - in case that
the controller goes down another node should take other (that;s for example
how master/slave
architecturere works if the master fails - a slave takes other).
Another solution would be to implement the same algorithm db clusters use
i.e. have a log for changes
and synchronize on that - no controller needed.
Actually the master/master cluster a very common for example in db world and
the algorithms for it
are well known.
Again ,i really appreciate the tomcat dev. work ,so if they do not have time
to implement such an alg.
then i understand.


> Now do you see why this hasn't been implemented?
>


No i don't .A complexity of solution does not mean "let's not do it". In
that case we should say:

"Hei, writing kernel is hard ,let's not do Linux"
or
"Hei,GC is hard let leave memory management in Java to programmers"

However,something like this is fine:

"We are short in resources and this feature is _not_ that needed ,let's do
something else."

And yes it _ is _ agreed that this feature is _ mostly _ not very needed so
if there is nobody to implement
it - let the developers do something else.

Evgeny


>
> - -chris
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.10 (MingW32)
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
>
> iEYEARECAAYFAkrl6L8ACgkQ9CaO5/Lv0PDdoACfR6+VP3xSV+17TdKs2759Z0fU
> gjAAnA5IDckHoO8AClD7Q1X+ArGl+Qte
> =0cmc
> -----END PGP SIGNATURE-----
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
> For additional commands, e-mail: users-h...@tomcat.apache.org
>
>

Reply via email to