[heketi-devel] Remove Device: Used to distribute all the bricks from device to other devices
Luis Pabon
lpabon at chrysalix.org
Fri Mar 10 06:46:22 UTC 2017
For 1. No, sounds good.
2. maybe both mean the same?
On Thu, Mar 9, 2017 at 5:14 PM, Raghavendra Talur <rtalur at redhat.com> wrote:
> On Fri, Mar 10, 2017 at 3:01 AM, Michael Adam <obnox at samba.org> wrote:
> > On 2017-03-09 at 16:17 -0500, Luis Pabon wrote:
> >> This looks great. Thanks for the state machine diagram.
> >
> > Indeed!
> >
> >> On Thu, Mar 9, 2017 at 11:06 AM, Raghavendra Talur <rtalur at redhat.com>
> >> wrote:
> >>
> >> > The PR now has the API changes requested.
> >> >
> >> > Before review, here is the current state diagram
> >> >
> >> >
> >> >
> >> > +-----------------+ disable/offline +------------------+
> >> > | |--------------------->| |
> >> > | | | Offline/Disabled |
> >> > | Online/Enabled |<---------------------| |
> >> > | | enable/online | |
> >> > +-----------------+ +------------------+
> >> > ^ ^ |
> >> > | | |remove
> >> > | offline| |
> >> > |add | |
> >> > | | |
> >> > | | |
> >> > | | v
> >> > +------------------+ +-------------------+
> >> > | | | |
> >> > | Deleted | | Failed/Removed |
> >> > | |<--------------------| |
> >> > | | delete | |
> >> > +------------------+ +-------------------+
> >> >
> >> >
> >> >
> >> > The current implementation *requires* the device to be in "Offline"
> state
> >> > before it can be removed. Some of the operations shown above aren't
> >> > implemented yet.
> >
> > I was wondering if some of those should or can
> > be implemented at all:
> >
> > * deleted ----[add]----> online
> > * failed/removed --[offline]--> offline/disabled
> >
> > I assume that these are the ops not implemented (yet).
> yes, that's right.
>
> >
> > I think remove should be a one-way operation.
> > And similar for deletion. Or is the 'deleted' state
> > something that is equivalent to a new disk not being
> > used at all in any volume?
> yes, I considered it as a new disk which can be added.
> But, on testing I learnt: as the device is not completely cleaned up
> on a delete, heketi does not consider it as *new* and add fails. Needs
> a little work here.
>
> regarding *failed*-->*offline* state
> current status: not implemented
> I propose that we should implement it.
>
> I have more questions:
> 1. Prasanth asked us to give cli aliases for enable/disable as
> online/offline. I would like to do it. Any objections?
> 2. What should be the state name, *failed* or *removed*? In the code,
> it is failed everywhere. The operation is called *remove*, but the
> device info output has *failed*. We really need to choose one of them
> here and be consistent.
>
> Raghavendra Talur
>
> >
> > Apart from these questions -- great!
> >
> > Thanks - Michael
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/heketi-devel/attachments/20170310/82dc0456/attachment.html>
More information about the heketi-devel
mailing list