Hi All,

The CAPI team is working on deprecating usage of nfs for the default
blobstore in cf-release in favor of the webdav protocol.

I'm interested to know if anyone is using their own separate nfs, not the
included nfs job, with cf-release. We are considering if it's feasible to
remove all support for NFS.

Please let me know if you have questions/concerns about this change.

This will help resolve some problems caused due to mount timing issues and
allow bosh to change bosh delete deployment to wait for processes to stop
before deleting vms. [1] [2] I believe it will also allow the bosh team to
further harden the stemcell.

For existing deployments using the included nfs job, this will require a
few new manifest entries, but otherwise should be seamless and not require
migration of blobs as we will simply start up a webdav server on the nfs
job.

Thanks,
-Dieu
CF CAPI PM

[1] https://www.pivotaltracker.com/story/show/99462742
[2] https://www.pivotaltracker.com/story/show/62354622

Hi Dieu,

We are currently directly using an NFS mount provided by our enterprise
NetApp appliance. We have been aware of this desired change but haven't
had time to investigate what we will do about it yet. I'm not sure if
NetApp provides a WebDav endpoint we can directly use or if we'll have to
create an NFS job that mounts in the WebDav job you will be providing.
We'd be grateful for any ideas or help in this change.

Mike

On Thu, Jan 14, 2016 at 11:20 PM, Dieu Cao <dcao(a)pivotal.io> wrote:

Hi All,

The CAPI team is working on deprecating usage of nfs for the default
blobstore in cf-release in favor of the webdav protocol.

I'm interested to know if anyone is using their own separate nfs, not the
included nfs job, with cf-release. We are considering if it's feasible to
remove all support for NFS.

Please let me know if you have questions/concerns about this change.

This will help resolve some problems caused due to mount timing issues
and allow bosh to change bosh delete deployment to wait for processes to
stop before deleting vms. [1] [2] I believe it will also allow the bosh
team to further harden the stemcell.

For existing deployments using the included nfs job, this will require a
few new manifest entries, but otherwise should be seamless and not require
migration of blobs as we will simply start up a webdav server on the nfs
job.

Thanks,
-Dieu
CF CAPI PM

[1] https://www.pivotaltracker.com/story/show/99462742
[2] https://www.pivotaltracker.com/story/show/62354622

Hi Mike,

We could keep the nfs mounting job and the capability for cc to connect to
an external non-bosh deployed nfs as an option.
We would still like to replace the nfs job with the webdav job as the
default for new deployments in cf-release if nothing else is configured.
Would that work for you?

-Dieu

On Fri, Jan 15, 2016 at 9:03 AM, Mike Youngstrom <youngm(a)gmail.com> wrote:

Hi Dieu,

We are currently directly using an NFS mount provided by our enterprise
NetApp appliance. We have been aware of this desired change but haven't
had time to investigate what we will do about it yet. I'm not sure if
NetApp provides a WebDav endpoint we can directly use or if we'll have to
create an NFS job that mounts in the WebDav job you will be providing.
We'd be grateful for any ideas or help in this change.

Mike

On Thu, Jan 14, 2016 at 11:20 PM, Dieu Cao <dcao(a)pivotal.io> wrote:

Hi All,

The CAPI team is working on deprecating usage of nfs for the default
blobstore in cf-release in favor of the webdav protocol.

I'm interested to know if anyone is using their own separate nfs, not
the included nfs job, with cf-release. We are considering if it's feasible
to remove all support for NFS.

Please let me know if you have questions/concerns about this change.

This will help resolve some problems caused due to mount timing issues
and allow bosh to change bosh delete deployment to wait for processes to
stop before deleting vms. [1] [2] I believe it will also allow the bosh
team to further harden the stemcell.

For existing deployments using the included nfs job, this will require a
few new manifest entries, but otherwise should be seamless and not require
migration of blobs as we will simply start up a webdav server on the nfs
job.

Thanks,
-Dieu
CF CAPI PM

[1] https://www.pivotaltracker.com/story/show/99462742
[2] https://www.pivotaltracker.com/story/show/62354622

Yes I think that should work fine.

Thanks,
Mike

On Fri, Jan 15, 2016 at 3:17 PM, Dieu Cao <dcao(a)pivotal.io> wrote:

Hi Mike,

We could keep the nfs mounting job and the capability for cc to connect
to an external non-bosh deployed nfs as an option.
We would still like to replace the nfs job with the webdav job as the
default for new deployments in cf-release if nothing else is configured.
Would that work for you?

-Dieu

On Fri, Jan 15, 2016 at 9:03 AM, Mike Youngstrom <youngm(a)gmail.com>
wrote:

Hi Dieu,

We are currently directly using an NFS mount provided by our enterprise
NetApp appliance. We have been aware of this desired change but haven't
had time to investigate what we will do about it yet. I'm not sure if
NetApp provides a WebDav endpoint we can directly use or if we'll have to
create an NFS job that mounts in the WebDav job you will be providing.
We'd be grateful for any ideas or help in this change.

Mike

On Thu, Jan 14, 2016 at 11:20 PM, Dieu Cao <dcao(a)pivotal.io> wrote:

Hi All,

The CAPI team is working on deprecating usage of nfs for the default
blobstore in cf-release in favor of the webdav protocol.

I'm interested to know if anyone is using their own separate nfs, not
the included nfs job, with cf-release. We are considering if it's feasible
to remove all support for NFS.

Please let me know if you have questions/concerns about this change.

This will help resolve some problems caused due to mount timing issues
and allow bosh to change bosh delete deployment to wait for processes to
stop before deleting vms. [1] [2] I believe it will also allow the bosh
team to further harden the stemcell.

For existing deployments using the included nfs job, this will require
a few new manifest entries, but otherwise should be seamless and not
require migration of blobs as we will simply start up a webdav server on
the nfs job.

Thanks,
-Dieu
CF CAPI PM

[1] https://www.pivotaltracker.com/story/show/99462742
[2] https://www.pivotaltracker.com/story/show/62354622

dieu, i thought i remember dmitriy indicating there are some packages he
wants to remove from the stemcell for NFS. would the "bring your own NFS
server" approach still allow the stemcell changes dmitriy is advocating?

On Fri, Jan 15, 2016 at 7:21 PM, Mike Youngstrom <youngm(a)gmail.com> wrote:

Yes I think that should work fine.

Thanks,
Mike

On Fri, Jan 15, 2016 at 3:17 PM, Dieu Cao <dcao(a)pivotal.io> wrote:

Hi Mike,

We could keep the nfs mounting job and the capability for cc to connect
to an external non-bosh deployed nfs as an option.
We would still like to replace the nfs job with the webdav job as the
default for new deployments in cf-release if nothing else is configured.
Would that work for you?

-Dieu

On Fri, Jan 15, 2016 at 9:03 AM, Mike Youngstrom <youngm(a)gmail.com>
wrote:

Hi Dieu,

We are currently directly using an NFS mount provided by our enterprise
NetApp appliance. We have been aware of this desired change but haven't
had time to investigate what we will do about it yet. I'm not sure if
NetApp provides a WebDav endpoint we can directly use or if we'll have to
create an NFS job that mounts in the WebDav job you will be providing.
We'd be grateful for any ideas or help in this change.

Mike

On Thu, Jan 14, 2016 at 11:20 PM, Dieu Cao <dcao(a)pivotal.io> wrote:

Hi All,

The CAPI team is working on deprecating usage of nfs for the default
blobstore in cf-release in favor of the webdav protocol.

I'm interested to know if anyone is using their own separate nfs, not
the included nfs job, with cf-release. We are considering if it's feasible
to remove all support for NFS.

Please let me know if you have questions/concerns about this change.

This will help resolve some problems caused due to mount timing issues
and allow bosh to change bosh delete deployment to wait for processes to
stop before deleting vms. [1] [2] I believe it will also allow the bosh
team to further harden the stemcell.

For existing deployments using the included nfs job, this will require
a few new manifest entries, but otherwise should be seamless and not
require migration of blobs as we will simply start up a webdav server on
the nfs job.

Thanks,
-Dieu
CF CAPI PM

[1] https://www.pivotaltracker.com/story/show/99462742
[2] https://www.pivotaltracker.com/story/show/62354622

--
Thank you,

James Bayer