[Haizea] Multiple virtual machines on single physical machine

[Borja Sotomayor]

Hi,

> I have done the same thing but i was getting error. So I have searched and found that one more correction is required which is in rpc_commands.py as shown below.
> Changed line :
>             capacity.set_quantity(constants.RES_CPU, int(self.opt.cpu * 100))
>
> Originally it was :
>             capacity.set_quantity(constants.RES_CPU, int(self.opt.cpu) * 100)

Ah, yes, that's correct. I've just committed a fix for this in the repository:

https://phoenixforge.cs.uchicago.edu/repositories/revision/haizea/793

Thanks for pointing it out!


> I have seen details of on which physical node each lease will be scheduled so i found that it should schedule 8th lease. To check this thing you will also have to do these two changes in rpc_commands.py file.First one which you told me and second one i have mentioned above.

If you go just by the requested CPU, the 8th lease should, indeed, be
scheduled (in particular, physical node 2 should still have 20% CPU
available for it). However, after the first seven leases have been
requested, node 2 (which has 1024MB of memory) has two VMs scheduled
on it, each requesting 512MB. So, there is no memory available for a
third VM. If I change the configuration file so that nodes have 2048MB
of memory, I can request the first 8 leases (but not the 9th one
which, as you point out, cannot be scheduled because there is no CPU
available for it).

btw, one way of debugging these problems is to switch the logging
level to VDEBUG (this is what I did to figure out what was wrong).
This will give you more detailed information about the scheduling
decisions that are taking place.

Cheers!
--Borja