Being a command line guy myself, I tend to agree and disagree. You can wrap up nearly all of your day-to-day tasks in a gui (be it X, web or other). Your gui may only provide 75% of the functionality of your software. The kind of thing i had in mind was more like, "hey, we got a new initrd//kernel, lets push that out to the pxeboot area and we'll do rolling reboots of the nodes" and get the cluster/web framework to manage live migration of hosts (or shutdown/startup) from one node to another. Plus things like, click, click, click - i just provisioned a new guest with a some disk space, a network connection and a chunk of memory/cpu. Of course the most important part of that is "how's the cluster going this morning" - *click* its all green, good.
That doesn't mean your free from the command, god no cause a gui limits (quite harshly at times) what your capable of doing - but it can simplify your life. But why web? <story>a while ago i was deploying a bunch of sun servers doing some web server work for a large corp (by AU terms, 50k employee's) mostly j2ee, iplanet web server and oracle. They used HP openview (among other things), but the op's guys jumped on and installed big brother agents. The guy said to me "oh yeah, we have this little server over here running bb for ourselves cause we can do almost anything and monitor almost anything with it. If its not in BB already its like easy enough to throw together a script of some kind to measure/monitor it. Basically, what i do is set the warning levels really low so i can just pull up my java nokia phone and go thru the wap portal and see if im likely to get paged before i go anywhere. If its yellow, OV wont alarm but I know I had best take a look before i go anywhere".</story> But, its always made me aim for a simple interface for day-to-day and get funky for the special cases, disaster recovery, etc. Now as the thing im trying to manage gets large, the better I want that interface to be because if i have to manage 4 machines i could ssh into them all and just look, but as that becomes 8 machines, 16, 32, etc the more i need information compressed or the harder its going to be for me to get to terms with it (this may come across as me sounding like im missing the point of kvmfs, which it inst intended to). But, i whole-heartedly agree with kvmfs, it looks brilliant and exactly what vm's really do need! Of course, prefix everything i just said with "in my humble opinion". ron minnich wrote: > On 7/31/07, Paul J R <[EMAIL PROTECTED]> wrote: > > >> The kind of thing I had in mind was relatively simple, it'd be a >> miniture OS that would boot up (over pxe/usb/cdrom) into ram (its about >> 90mb of ram so far with kvm and a full kernel + kudzu and few small >> components such as a httpd+php). it searches for a configuration >> partition and away you go managing the rest via the web interface. I had >> clustering/multi-node sitting at the back of my head as well. >> > > I can tell you that for clusters web console interfaces are pretty > useless, as scripting is hard for them, and a web interface is really > pretty clumsy for almost anything of reasonable size. I think that > kvmfs is really a great idea, and what I would want if I had a 128 or > 1024 node or larger cluster. Note that we plan9 to do sims here with > 10,000 instances of lguest on a 128-node cluster, and a web interface > would be the worst possible way to do that. > > ron > > ------------------------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/ _______________________________________________ kvm-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/kvm-devel
