On Fri, Apr 6, 2012 at 2:41 PM, Sami Tikka <[email protected]> wrote:
> So what you are saying is that hard disk is becoming the bottleneck
> for you if there are multiple builds running at the same time even if
> you have free slaves picking their nose?
>
> It sounds to me like you can solve your problem by setting number of
> executors in each slave to 1.
>
> At work I have 20 slaves each with 1 executor and the builds REALLY ROAM :)
No, he is saying that after jobs have been run, subsequent runs will
queue waiting for the node that did the previous build instead of
migrating to an available node even if it is a long wait. I think if
that node is down or disconnected, the job would go to a different one
immediately, but if it is just busy the job stays in the queue for it
no matter how deep the queue gets. In the case of a big source
checkout with a small update, it might be worth the wait but I don't
know if there is any way to control the behavior when that's not the
case.
--
Les Mikesell
[email protected]
