users@glassfish.java.net

Re: capturing redeployment downtime -- cluster?

From: Hong Zhang <Hong.Zhang_at_Sun.COM>
Date: Mon, 07 Jul 2008 09:53:58 -0400

Hi, Kristian
   You can take a look at the rolling upgrade doc here:

http://docs.sun.com/app/docs/doc/819-3679/abdik?l=en&a=view&q=rolling


- Hong

Kristian Rink wrote:

>Folks;
>
>currently experiencing the (obvious) situation that, as soon as a webapp
>gets (re)deployed to the server, it is not available to external users
>(who by then will just be provided with a not-that-friendly error page).
>I am searching for a way of doing "seamless deployment", i.o.w.
>(re)deploying applications without the users noticing this. Simple idea,
>initially, clustering comes to my mind, considering at least two
>glassfish nodes running the same applications, having always "one
>running" while the other one gets a new application version deployed,
>providing users with an uninterrupted service. Question coming to my
>mind about this is: What's the best way of doing so? Build a cluster
>running two nodes, manually (and sequentially) deploying the
>applications to node#1 and node#2, letting pass a reasonable amount of
>time in between in order to have sessions replicated across the nodes?
>Or does glassfish a more sophisticated means of implementing such a
>scenario? Given apache2/mod_proxy_balancer in front of it, at least the
>balancing / redirection of HTTP requests to "only" the working node can
>be easily enforced so to me it comes down to meaningfully configure
>glassfish.
>
>Ideas, anyone?
>TIA and best regards,
>Kristian
>
>
>