Which hypervisor is this? Are you saying that you have 4 spindles and you want to dedicate one spindle per VM (as an example).
On 6/4/13 12:17 PM, "David Ortiz" <dpor...@outlook.com> wrote: >Andrew, > I was able to tag my nodes so that I could ensure that my hadoop >nodes would deploy to the same compute node as their storage. I also was >able to use a network topology script to make sure that the data is >distributed among two hosts at minimum. The local storage pool thing is >what is getting me at the moment. To use my available local storage I >had to set it up as one volume for cloudstack to be able to use all of >it, but as a result I have that serving as a bottleneck for the three >slave nodes on each host since it is already slowed down by the extra >abstraction of the drives, and has three vms trying to slam it at once. >Thanks, David Ortiz > >> From: andrew.ba...@gmail.com >> Date: Tue, 4 Jun 2013 12:05:32 -0700 >> Subject: Re: Hadoop cluster running in cloudstack >> To: users@cloudstack.apache.org >> >> This is a very interesting question - I know I've asked for better >>control >> over where VMs end up going, specifically to be able to ensure rack >> locality for Hadoop nodes, but I don't know what the progress has been >>on >> that, nor do I know whether there's a way of doing multiple storage >>pools >> on a single host short of some really silly jumping through hoops, like >> running multiple cloud agents on a single host. But I could be wrong - >>I'm >> not as in touch with the internals as others may be. >> >> A. >> >> On Tue, Jun 4, 2013 at 11:43 AM, David Ortiz <dpor...@outlook.com> >>wrote: >> >> > Hello, >> > Has anyone tried running a hadoop cluster in a cloudstack >>environment? >> > I have set one up, but I am finding that I am having some IO >>contention >> > between slave nodes on each host since they all share one local >>storage >> > pool. As I understand it, there is not currently a method for using >> > multiple local storage pools with VMs through cloudstack. Has anyone >>found >> > a workaround for this by any chance? >> > Thanks, David Ortiz >