The big gain is that we're able to gain all the perks of a framework with just the overhead of running a container which we run all via marathon. I've been able to scale it locally on our setup a bit and have ES report the additional slaves are healthy.
On Mon, Sep 15, 2014 at 10:11 AM, CCAAT <[email protected]> wrote: > On 09/15/14 08:30, Adam Shannon wrote: > >> Hello all, >> >> I recently was able to setup what the subject says, and was wondering if >> there's anyone else doing something similar. (Or if there are other >> ideas around this subject.) >> >> Basically, I took the official mesos framework, forked and updated >> versions for the versions that I'm running. You can see the tree here: >> https://github.com/Banno/elasticsearch-mesos/tree/cleanup >> >> It's then easy to build a container to extract and run the framework. >> https://github.com/Banno/docker-elasticsearch-mesos Which will then just >> work with mesos to scale / manage itself. >> >> The tricky thing is that because this is just a container marathon can >> run it. Marathon can scale it, and manage running it on whatever slave. >> >> Thoughts? >> > > what is the advantage to this approach/configuration? > I'm new to mesos and from what I've read about I do not > see a lot of details of what/how folks hope to solve problems. > > My area of interest is Big Science. More specifically, using > mesos + spark + ??? to run simulations of CO2 injected into the ground. > So that means billions of PDE (partial differential equations)/matrix > calculations/solutions on a 3D geometry. Keeping it all "in_memory" > would be keen. > > > Adam Shannon | Software Engineer | Banno | ProfitStars® >> 206 6th Ave Suite 1020 | Des Moines, IA 50309 | Cell: 515.867.8337 >> <tel:515.867.8337> >> > > -- Adam Shannon | Software Engineer | Banno | ProfitStars® 206 6th Ave Suite 1020 | Des Moines, IA 50309 | Cell: 515.867.8337
