Re: Recreate strategy optimization

[Clayton Coleman]

That's not possible today, although in the future I could imagine
support for that.  There would have to be some gating factor that
prevents pod startup which is not yet possible.

You *can* pre-seed a cluster with images by creating a daemon set in
your prestart deployment hook that contains the image, then wait until
enough pods have started to continue.  The scheduler would then prefer
those nodes because your image is already pulled.

> On Aug 8, 2016, at 4:52 PM, Philippe Lafoucrière 
>  wrote:
>
> Hi,
>
> I have noticed something that could be optimized in recreate strategy: During 
> a deploy, Openshift will set the current running replicas to zero, and then 
> pull the new image, and eventually starts it.
> The problem is, sometimes, the pull can very (very) long, and during that 
> period, the pods are unavailable. I would be more efficient to start pulling 
> the image, and IF the image is pulled successfully, start to touch replicas.
> What do you think?
>
> Thanks,
> Philippe
> ___
> users mailing list
> users@lists.openshift.redhat.com
> http://lists.openshift.redhat.com/openshiftmm/listinfo/users

___
users mailing list
users@lists.openshift.redhat.com
http://lists.openshift.redhat.com/openshiftmm/listinfo/users

Recreate strategy optimization

[Philippe Lafoucrière]

Hi,

I have noticed something that could be optimized in recreate strategy:
During a deploy, Openshift will set the current running replicas to zero,
and then pull the new image, and eventually starts it.
The problem is, sometimes, the pull can very (very) long, and during that
period, the pods are unavailable. I would be more efficient to start
pulling the image, and IF the image is pulled successfully, start to touch
replicas.
What do you think?

Thanks,
Philippe
___
users mailing list
users@lists.openshift.redhat.com
http://lists.openshift.redhat.com/openshiftmm/listinfo/users