Re: [Condor-users] Can CONDOR emulate resources

[Brian Candler <B.Candler@xxxxxxxxx>]

On Tue, Aug 14, 2012 at 06:48:45AM +0000, Yuval Leader wrote:
>    1.      I would like to use CONDOR with a single server emulating many
>    servers, for testing purposes. Since I do not have all the execute
>    machine yet, I would like one real server to mimic the behavior of the
>    daemons running on the execution machines, for many machines. By “many”
>    I target 1000-5000 servers. What is the best way to do it ?
> 
>    2.      The per-execute machine resource attributes (such as cpu,
>    memory, state etc) is now stored in a file. How do I take this data and
>    send it to the condor pool as if it came from condor_startd running on
>    the real servers ?

Run multiple instances of condor_startd?

http://research.cs.wisc.edu/condor/manual/v7.8/3_3Configuration.html

"This syntax has been further expanded to allow for the specification of a
local name on the command line using the command line option

  -local-name <local-name>

This allows multiple instances of a daemon to be run by the same
condor_master daemon, each instance with its own local configuration
variable."

Whether 1000-5000 separate startd instances is realistic I don't know.

Taking this further, you could have 1000 VMs each running Condor.  This
probably doesn't scale for full virtualisation, but it might using something
like LXC or OpenVZ.

>    3.      Is there a method to request a future job execution e.g. run on
>    a required resource starting in 2 weeks for 1 day, and still get a
>    confirmation now that the resource will be indeed claimed.

AFAIK there's no way of doing advance reservations. The job sits in the
queue until some machine is ready to take it.  You can set up your job
ClassAd so that it *won't* execute except in a time window that you choose,
but when that time comes along, it'll have to fight with all others.