Hey Matt: I have a question regarding how federation works.
So let's say I have two "clusters" each with 100 nodes. Each node is running gmond and on the headnode of each cluster, I run gmetad on it. When I connect to the gmetad port on each headnode, it is wrapped in a <GRID> tag (scalability mode is enabled by default) and it has metrics of _all_ the compute nodes in the cluster. I have another server which I would like to monitor both clusters, so I have gmetad and the webfrontend installed and the data_sources point to the gmetad ports of both clusters. I can then see the summary in the frontpage of the webfrontend and if needed I can drill down to see more fine-grain detail. So everything is working fine and dandy, however, my question is -- when my "master" server queries each cluster's gmetad port, it will get back the _entire_ XML stream with all the compute nodes information, correct? If I then connect to my "master" server's gmetad port, I will not get all the detail of the compute nodes, but just a summary (since that data is filtered). Is my understanding of how the federation works correct? Thanks in advance, Bernard On 6/13/07, Matt Massie <[EMAIL PROTECTED]> wrote: > http://ganglia.info/papers/Sacerdoti03Monitoring.pdf > > is the paper which details federation. > > if you look at the example gmetad.conf in the distribution, you'll find > comment about "scalability mode". this allows you to use a hierarchical > delegation model (federation) which is extremely scalable. it's almost > as simple as adding... > > scalable on > > to your gmetad.conf file. almost. > > -matt > > Buccaneer for Hire. wrote: > > Is there a paper somewhere explaining federation in > > detail? > > > > I am trying to build a monitor station anyone can come > > to see what is going on with the clusters/grids that > > we are responsible for, but I am having problems > > trying to get the thing layed out. > > > > My problem is: > > > > (1) We manage a large cluster (2500+nodes/400+TB) and > > now have it monitoring just fine for now and reporting > > to two nodes. We are monitoring by scheuling queue > > and infrastructure. > > > > (2) I have a grid composed of multiple clusters and > > workstations with all the information coming to two > > nodes/cluster and forwarding to the grid head node > > (also bing . Some cluster can be resized dynamically > > at any time. > > > > (3) a number of other smaller cluster (32 nodes or > > so)in differing locations. > > > > I want to federate this to the uber monitor so that I > > get summaries on the front page and can then drill > > down. On this big cluster, it means I don't want all > > 10 queues showing at the bottom until I start to drill > > down. Does that make any sense to anyone? > > > > TIA, > > > > > > > > > > ____________________________________________________________________________________ > > Looking for a deal? Find great prices on flights and hotels with Yahoo! > > FareChase. > > http://farechase.yahoo.com/ > > > > ------------------------------------------------------------------------- > > This SF.net email is sponsored by DB2 Express > > Download DB2 Express C - the FREE version of DB2 express and take > > control of your XML. No limits. Just data. Click to get it now. > > http://sourceforge.net/powerbar/db2/ > > _______________________________________________ > > Ganglia-general mailing list > > [email protected] > > https://lists.sourceforge.net/lists/listinfo/ganglia-general > > > > > > > > > > ------------------------------------------------------------------------- > This SF.net email is sponsored by DB2 Express > Download DB2 Express C - the FREE version of DB2 express and take > control of your XML. No limits. Just data. Click to get it now. > http://sourceforge.net/powerbar/db2/ > _______________________________________________ > Ganglia-general mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/ganglia-general > ------------------------------------------------------------------------- This SF.net email is sponsored by DB2 Express Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now. http://sourceforge.net/powerbar/db2/ _______________________________________________ Ganglia-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/ganglia-general
