Re: Persistent Volumes on Cloud Foundry

James Bayer


if you look at this section of the doc [1] it discusses "Reattachable
Volumes" which are similar to EBS volumes attaching to EC2 instances
without allowing multiple instances to be bound to the same volume at the
same time. that likely aligns better with a database style use case
(although performance of remote volumes will certainly be a consideration
depending on the performance requirements). there is some new diego
scheduling work required for "Reattachable Volumes".

"Distributed Filesystem" (same volume attached to multiple instances at the
same time like NFS) and "Scratch" (temporary extra disk) are use cases
which are planned to be supported first as they do not require diego
scheduler changes to use now. the doc also references snapshots as
something that may be needed for this use case. there is discussion going
on in the doc if you want to continue it there.


On Tue, Apr 5, 2016 at 8:02 PM, Dr Nic Williams <drnicwilliams(a)>

Is a goal of this work to run pure data services like PostgreSQL or Redis
inside "application instances"?

On Tue, Apr 5, 2016 at 5:01 PM -0700, "Ted Young" <tyoung(a)>

This proposal describes the changes necessary to the Service Broker API to
utilize the new Volume Management features of the Diego runtime. It
contains examples of how services can provide a variety of persistent data
access to CF applications, where each service maintains control of the
volume lifecycle. This is to allow some services that provide blank storage
space to applications, and other services that provide access to complex or
externally managed data (such as IBM's Watson).

We are moving fast on delivering a beta of this feature, so please have a
look and give feedback now if this is of interest to you. More detail will
be added to the proposal as necessary.


Ted Young
Senior Engineer / Product Manager
Pivotal Cloud Foundry

Thank you,

James Bayer

Join { to automatically receive all group messages.