A nice enhancement would be automatically discovering other voldemort instances
and downloading a shared config. This would alleviate the config file issue as
there would just be a bunch of shared settings. For example, when starting a
new node you just need the location of one server and once connected, it would
locate all of the other servers on the network. This means that you wouldn't
have to maintain a list of all servers on every host.
Also, the two config files that are static on every host (cluster.xml and
stores.xml) would be "downloaded" from the other servers. All of the
information would be stored in the database.
This would make the cluster much more dynamic as spinning up a new instance
would instantly connect and start doing work (after downloading its share of
data). You wouldn't have to update all of the other machines configs.
In the per-node configuration you would just have to list a number of seed
machines so that it can connect and find the others.
Original issue reported on code.google.com by kevincox...@gmail.com on 23 Feb 2013 at 6:59
Original issue reported on code.google.com by
kevincox...@gmail.com
on 23 Feb 2013 at 6:59