Re: Follow up on providing `--reconfiguration_policy=any` in future versions

[Benno Evers]

So, I basically see two options to getting a more permissive
reconfiguration policy committed:

First, implement the 'any' option as shown in the review, and add some
warnings to the documentation that things might break if it is used too
liberally.

Or second, add another option that allows arbitrary changes of  resources,
attributes etc. but still requires some safety constraints (e.g. hostname
must stay the same)

Personally, I would prefer the first, since I still feel like the second
option will over-complicate the situation for little tangible benefit.

Vinod, what do you think?

On Wed, May 9, 2018 at 5:45 PM, Zhitao Li  wrote:

> Hi Benno,
>
> Now that 1.6 RC is out, I think we have another 3-4 months and I'd like to
> get a resolution on whether we will make this happen in Mesos, and details
> of how it happened.
>
> I put some comments on MESOS-8600 about whether we should overwrite
> checkpointed data upon permitted changed.
>
>
> On Thu, Feb 22, 2018 at 9:44 AM, Benno Evers 
> wrote:
>
> > Hi,
> >
> > I'm afraid these were mostly informal lunch-time discussions, but I just
> > opened https://issues.apache.org/jira/browse/MESOS-8600 to have a place
> > where we can record any progress on this issue.
> >
> > Best regards,
> > Benno
> >
> > On Tue, Feb 20, 2018 at 7:47 PM, Zhitao Li 
> wrote:
> >
> > > Hi Benno,
> > >
> > > Thanks for the the diff. I took a quick look and cannot anticipate
> issues
> > > with it in our environment. I'll talk to our release manager to try it
> > out
> > > in our environment.
> > >
> > > w.r.t. to the health check issue, is there meeting notes or a JIRA
> issue
> > > capturing existing discussions?
> > >
> > > On Mon, Feb 19, 2018 at 9:59 AM, Benno Evers 
> > > wrote:
> > >
> > > > Hi Zhitao,
> > > >
> > > > great to see that there's interest in this.
> > > >
> > > > The most specific concern that we had at the time was that we were
> not
> > > > sure about the best way to handle health checks on agents where
> > > > the hostname changed. (together with a general feeling
> > > > that we needed a bit more time to think through possible failure
> > > > scenarios)
> > > >
> > > > If you're willing to blaze trail, you could apply
> > > >
> > > > https://reviews.apache.org/r/64384/
> > > >
> > > > and see if this causes any observable issues.
> > > >
> > > > Of course, I'm also up for a follow-up meeting.
> > > >
> > > > Best regards,
> > > > Benno
> > > >
> > > >
> > > > On Thu, Feb 15, 2018 at 9:03 PM, Zhitao Li 
> > > wrote:
> > > >
> > > > > Hi Vinod/Benno,
> > > > >
> > > > > This is a follow up from MESOS-1739. We have recently discovered
> some
> > > > > previous unknown use cases which a fully allowed
> > > > > `--reconfiguration_policy-any` from the design doc
> > > > >  > > > > KxwU4lLtr53SrE5U3Q/edit#>
> > > > > will
> > > > > really help our operation.
> > > > >
> > > > > Do we want to have a follow up meeting to see what's blockers to
> > fully
> > > > > implement that?
> > > > >
> > > > > Thanks.
> > > > >
> > > > >
> > > > > --
> > > > > Cheers,
> > > > >
> > > > > Zhitao Li
> > > > >
> > > >
> > > >
> > > >
> > > > --
> > > > Benno Evers
> > > > Software Engineer, Mesosphere
> > > >
> > >
> > >
> > >
> > > --
> > > Cheers,
> > >
> > > Zhitao Li
> > >
> >
> >
> >
> > --
> > Benno Evers
> > Software Engineer, Mesosphere
> >
>
>
>
> --
> Cheers,
>
> Zhitao Li
>



-- 
Benno Evers
Software Engineer, Mesosphere

Re: Follow up on providing `--reconfiguration_policy=any` in future versions

[Zhitao Li]

Hi Benno,

Now that 1.6 RC is out, I think we have another 3-4 months and I'd like to
get a resolution on whether we will make this happen in Mesos, and details
of how it happened.

I put some comments on MESOS-8600 about whether we should overwrite
checkpointed data upon permitted changed.


On Thu, Feb 22, 2018 at 9:44 AM, Benno Evers  wrote:

> Hi,
>
> I'm afraid these were mostly informal lunch-time discussions, but I just
> opened https://issues.apache.org/jira/browse/MESOS-8600 to have a place
> where we can record any progress on this issue.
>
> Best regards,
> Benno
>
> On Tue, Feb 20, 2018 at 7:47 PM, Zhitao Li  wrote:
>
> > Hi Benno,
> >
> > Thanks for the the diff. I took a quick look and cannot anticipate issues
> > with it in our environment. I'll talk to our release manager to try it
> out
> > in our environment.
> >
> > w.r.t. to the health check issue, is there meeting notes or a JIRA issue
> > capturing existing discussions?
> >
> > On Mon, Feb 19, 2018 at 9:59 AM, Benno Evers 
> > wrote:
> >
> > > Hi Zhitao,
> > >
> > > great to see that there's interest in this.
> > >
> > > The most specific concern that we had at the time was that we were not
> > > sure about the best way to handle health checks on agents where
> > > the hostname changed. (together with a general feeling
> > > that we needed a bit more time to think through possible failure
> > > scenarios)
> > >
> > > If you're willing to blaze trail, you could apply
> > >
> > > https://reviews.apache.org/r/64384/
> > >
> > > and see if this causes any observable issues.
> > >
> > > Of course, I'm also up for a follow-up meeting.
> > >
> > > Best regards,
> > > Benno
> > >
> > >
> > > On Thu, Feb 15, 2018 at 9:03 PM, Zhitao Li 
> > wrote:
> > >
> > > > Hi Vinod/Benno,
> > > >
> > > > This is a follow up from MESOS-1739. We have recently discovered some
> > > > previous unknown use cases which a fully allowed
> > > > `--reconfiguration_policy-any` from the design doc
> > > >  > > > KxwU4lLtr53SrE5U3Q/edit#>
> > > > will
> > > > really help our operation.
> > > >
> > > > Do we want to have a follow up meeting to see what's blockers to
> fully
> > > > implement that?
> > > >
> > > > Thanks.
> > > >
> > > >
> > > > --
> > > > Cheers,
> > > >
> > > > Zhitao Li
> > > >
> > >
> > >
> > >
> > > --
> > > Benno Evers
> > > Software Engineer, Mesosphere
> > >
> >
> >
> >
> > --
> > Cheers,
> >
> > Zhitao Li
> >
>
>
>
> --
> Benno Evers
> Software Engineer, Mesosphere
>



-- 
Cheers,

Zhitao Li

Re: Follow up on providing `--reconfiguration_policy=any` in future versions

[Benno Evers]

Hi,

I'm afraid these were mostly informal lunch-time discussions, but I just
opened https://issues.apache.org/jira/browse/MESOS-8600 to have a place
where we can record any progress on this issue.

Best regards,
Benno

On Tue, Feb 20, 2018 at 7:47 PM, Zhitao Li  wrote:

> Hi Benno,
>
> Thanks for the the diff. I took a quick look and cannot anticipate issues
> with it in our environment. I'll talk to our release manager to try it out
> in our environment.
>
> w.r.t. to the health check issue, is there meeting notes or a JIRA issue
> capturing existing discussions?
>
> On Mon, Feb 19, 2018 at 9:59 AM, Benno Evers 
> wrote:
>
> > Hi Zhitao,
> >
> > great to see that there's interest in this.
> >
> > The most specific concern that we had at the time was that we were not
> > sure about the best way to handle health checks on agents where
> > the hostname changed. (together with a general feeling
> > that we needed a bit more time to think through possible failure
> > scenarios)
> >
> > If you're willing to blaze trail, you could apply
> >
> > https://reviews.apache.org/r/64384/
> >
> > and see if this causes any observable issues.
> >
> > Of course, I'm also up for a follow-up meeting.
> >
> > Best regards,
> > Benno
> >
> >
> > On Thu, Feb 15, 2018 at 9:03 PM, Zhitao Li 
> wrote:
> >
> > > Hi Vinod/Benno,
> > >
> > > This is a follow up from MESOS-1739. We have recently discovered some
> > > previous unknown use cases which a fully allowed
> > > `--reconfiguration_policy-any` from the design doc
> > >  > > KxwU4lLtr53SrE5U3Q/edit#>
> > > will
> > > really help our operation.
> > >
> > > Do we want to have a follow up meeting to see what's blockers to fully
> > > implement that?
> > >
> > > Thanks.
> > >
> > >
> > > --
> > > Cheers,
> > >
> > > Zhitao Li
> > >
> >
> >
> >
> > --
> > Benno Evers
> > Software Engineer, Mesosphere
> >
>
>
>
> --
> Cheers,
>
> Zhitao Li
>



-- 
Benno Evers
Software Engineer, Mesosphere

Re: Follow up on providing `--reconfiguration_policy=any` in future versions

[Zhitao Li]

Hi Benno,

Thanks for the the diff. I took a quick look and cannot anticipate issues
with it in our environment. I'll talk to our release manager to try it out
in our environment.

w.r.t. to the health check issue, is there meeting notes or a JIRA issue
capturing existing discussions?

On Mon, Feb 19, 2018 at 9:59 AM, Benno Evers  wrote:

> Hi Zhitao,
>
> great to see that there's interest in this.
>
> The most specific concern that we had at the time was that we were not
> sure about the best way to handle health checks on agents where
> the hostname changed. (together with a general feeling
> that we needed a bit more time to think through possible failure
> scenarios)
>
> If you're willing to blaze trail, you could apply
>
> https://reviews.apache.org/r/64384/
>
> and see if this causes any observable issues.
>
> Of course, I'm also up for a follow-up meeting.
>
> Best regards,
> Benno
>
>
> On Thu, Feb 15, 2018 at 9:03 PM, Zhitao Li  wrote:
>
> > Hi Vinod/Benno,
> >
> > This is a follow up from MESOS-1739. We have recently discovered some
> > previous unknown use cases which a fully allowed
> > `--reconfiguration_policy-any` from the design doc
> >  > KxwU4lLtr53SrE5U3Q/edit#>
> > will
> > really help our operation.
> >
> > Do we want to have a follow up meeting to see what's blockers to fully
> > implement that?
> >
> > Thanks.
> >
> >
> > --
> > Cheers,
> >
> > Zhitao Li
> >
>
>
>
> --
> Benno Evers
> Software Engineer, Mesosphere
>



-- 
Cheers,

Zhitao Li