On Fri, Jun 24, 2011 at 9:06 PM, Paul Graydon <p...@paulgraydon.co.uk>wrote:
> Hey all, > > Some collective brain picking if I can. We've been uhhming and ahhing > about the way we currently host our services. For the most part moving our > apps to the cloud is entirely impractical for various reasons. > Infrastructure that most need to access would never be permitted to be > accessed from outside of accessible and physically auditable area. > > The bulk of our apps are java apps, running from individual tomcat services > on our servers. I assume the reason for that is a failure of any individual > application doesn't take down multiple apps, though it does mean we waste a > fair bit of memory (albeit that's a cheap resource). > A number of them are coded badly, IMO, and can only be run in one location > at a time with app level cron processes and the like that cause serious > headaches. Recoding isn't a feasible option, or so I'm told every time I > suggest it. At the moment if a server dies I'm going to have to do a fair > bunch of manual work to bring up the apps on other servers and make sure all > is good. > > What I'm wondering about are what my options might be for Tomcat or similar > Platform as a Service clustering solution, and curious if there are anything > suitable? > > It's possibly a total pipe dream but I'd like to abstract away from the > hardware/OS as much as possible, and the more intelligent and self-healing > the better. I could go down the straight VMWare/Xen IaaS route internally. > On my pipe-dream list is to get puppet/chef config management in place, and > I could code something tied into monitoring to automatically trigger > failovers in case of hardware failure, or even smarter to automatically > juggle tomcats based on server & applet load. I'd like to pool my options > and see if there are more suitable alternatives, though. No point > duplicating effort if it's already been done! > Free is naturally desirable, and Open Source ideal, but its not an absolute > requirement. > > I'm fairly new to the java hosting world, this is the first place I've > worked that uses them. I must admit I've steadily gone from 'Urgh > black-box' to 'Wooo black-box!' regarding Tomcat. When things start going > weird in a tomcat instance its nice to be able to take the Windows route and > just turn it off and on again, and then pass on any stack-traces to devs to > investigate/fix. Especially given most of our tomcat apps start up in <5s. > > It looks like you're running pretty standard Java apps on top of Tomcat. Have you looked at Linux HA solutions to improve your availability in case a hardware node fails? We have a few servers at work in this configuration and it fails over the resources to the passive node just fine. -- Giovanni Tirloni
_______________________________________________ Tech mailing list Tech@lists.lopsa.org https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
