Hi Mark, I've updated the Github issue[1], it now includes a reference to a changeset[2] which does the following:
* Adds the ability for a node to create connections on the fly rather than have them pooled up front. This is done on a per-node basis, so technically you can have some nodes pooling and others not pooling (not sure why you'd want to do that though!). * Adds a sample config to the live tests which uses the on-the-fly connections. * Adds a load test which hits the test cluster using the on-the-fly connection option. For a given node, just set the *poolSize* value to 0 (zero) in the node's configuration, this will tell the node to operate in "create connections on the fly" mode. I'm guessing you'll want to do this for all nodes in your cluster. This should work with both cluster and proxy configurations. Feedback appreciated. OJ [1]: https://github.com/DistributedNonsense/CorrugatedIron/issues/111 [2]: https://github.com/DistributedNonsense/CorrugatedIron/commit/bbb641e4d761d714eebdb99e64dfe7ad2728989d On Tue, Mar 19, 2013 at 9:38 AM, OJ Reeves <o...@buffered.io> wrote: > Hello Mark, > > There is currently no implementation in CI which allows this. Having said > that, it's something that we have pondered ourselves. It might just be high > time to get that sorted out. > > I have created a github issue[1] so we can talk about it. May I request > that we move the discussion there? I am very keen to help get a solution > together for you. > > Thanks for the contact! > OJ > > [1]: https://github.com/DistributedNonsense/CorrugatedIron/issues/111 > > > On Tue, Mar 19, 2013 at 7:22 AM, Mark Morga <mmo...@rackspace.com> wrote: > >> TL;DR version: is there a way to use Corrugated Iron making connections on >> the fly rather than a fixed connection pool? >> >> Explanation: >> >> We are using Corrugated Iron on several (5) components that talk to Riak. >> We want to be able to independently scale the number of instances of these >> components live in production and we're having problems understanding how >> best to do this with the connection pool assigned in CI. >> >> Our Riak instance (currently 5 nodes) is load balanced behind an HAProxy >> instance and we have been using the RiakExternalLoadBalancer class pointed >> to this load balancer to instantiate our RiakClient. HAProxy is configured >> at the moment to permit up to 250 connections to each Riak node (we've >> played with this a bit). >> >> My concern is that as we pop additional nodes running one of our >> components, each of those will want to consume _n_ connections out of the >> available 250 and eventually we are going to run out. >> >> Looking through the code a bit today, I don't see how to use the Riak >> connections outside of either RiakExternalLoadBalancer or RiakCluster. >> Since RiakNode and RiakConnection are internal classes I don't see an easy >> way to do this. Ideas? >> >> Thanks in advance! >> >> Mark Morga >> >> >> _______________________________________________ >> riak-users mailing list >> riak-users@lists.basho.com >> http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com >> > > > > -- > > OJ Reeves > +61 431 952 586 > http://buffered.io/ > -- OJ Reeves +61 431 952 586 http://buffered.io/
_______________________________________________ riak-users mailing list riak-users@lists.basho.com http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
