►
Description
Kubernetes Storage Special-Interest-Group (SIG) Object Bucket Standup Meeting - 28 September 2020
Meeting Notes/Agenda: -
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
B
Okay,
so
this
is
the
last
week
we
have
before
the
cap
deadline
is
so
this
friday,
I
believe,
is
that
right,
srini.
B
Okay,
yeah
so
before
this
friday,
we'll
have
to
get
the
cap
in
we've
made
all
the
updates
as
per
the
comments
that
were
left
last
week
again
for
those
of
you
on
this
call,
please
take
some
time
to
look
at
the
cap
and
leave
your
feedback
on
it.
B
B
Yes,
that's.
B
Yeah,
we'll
only
have
one
more
review
before
this,
so
before
the
october
6
deadline.
I
think
it's
good
to
aim
to
finish
it
by
this
week
because
in
case
something
needs
to
be
changed.
We
can
use
the
next
week,
monday
and
tuesday
as
a
buffer.
B
So
yeah,
so
so,
our
other
big,
a
milestone
that
we're
working
towards
is
the
demo.
B
B
That
is,
as
the
code
is
built
in
such
a
way
that
it's
production
quality
where
it
is
tested
where
it
has
a
unit
test,
functional
tests
in
place,
and
it's
not
just
a
prototype
level
code
and
the
demo
that
we
want
to
show
as
the
first
milestone
is
a
very
simple
use
case
of
provisioning,
a
green
field
bucket
and
then
provisioning
a
bucket
access
to
gain
access
to
the
bucket
and
mounting
that
bucket
into
the
pod,
which
essentially
means
providing
the
bucket
name
and
the
credentials
to
the
part
that
wants
to
consume
it.
B
B
B
Here
are
the
tasks
that
we
wanted
to
accomplish,
but
this
is
the
all
the
tasks
that
needs
to
be
done.
This
this
includes
starts
that
needs
to
be
done
even
after
the
demo,
but
for
the
sake
of
the
demo
itself,
here
is
where,
as
we
as
of
last
time,
we
checked
give
me
one.
Second,
as
of
the
last
time
I
checked,
the
items
marked
in
green
are
the
items
that
we
have
already
completed.
B
So
what
I'm
going
to
do
today
is
check
the
progress
of
the
different
items
that
are
left
figure
out
if
there
are
any
blockers
and
try
to
resolve
them.
So,
let's
start
with
central
controller
manager.
B
So
one
of
the
things
I
want
to
do
with
with
this
development
is
kind
of
get
the
boilerplate
for
these
different
components
down
in
the
initially
so
have
the
framework
for
cli
set
in
place
where,
if
you
want
to
make
some
changes
like
add
a
flag
or
remove
a
flag,
it
should
be
very
simple:
we
shouldn't
be
dealing
with
the
creation
of
the
cli
after
the
logic
is
written.
B
That
would
just
make
things
a
little
complicated
to
fix.
So
I
want
to
start
with
this
simple
task,
so
I
believe
srini
is
working
on
this
shrini.
B
Have
you
I
haven't
looked
at
the
latest
code,
but
is
there
a
cli
implementation
for
the
central
controller
so.
C
B
Oh
before
we
go
into
that,
so
the
cli
is
in
use,
so
you're
saying
we
using
cobra
following
the
standard
that
we've
been
following
at
the
other.
B
Okay,
okay
and
then
you're
saying
something
about
bucket
request,
create
so
a
simple
bucket
request.
C
B
C
B
B
Okay,
what
what
kind
of
validation
are
you
doing
right
now.
E
C
Okay,
so
I
don't
remember
completely,
but
basic
field
validation
like
a
protocol
is
provided
that
kind
of
checks
that
I'm
doing.
Okay,.
B
B
C
C
All
of
them
I
have,
I
don't,
have
the
code
in
hand
right
now,
but
yeah.
There
are
some
checks,
but
it
creates
a
bucket
and
populates
with
the
required
fields,
and
I
can
I
can
check
the
bucket
object.
B
Okay,
input
field
validation,
like
bucket
name
bucket
name
format.
Check,
is
not
in
place
yet
yeah,
okay,
okay,
after
validation
on
bucket
request,
create
what
is
the
next
step?
You
do.
C
C
Parameters
and
all
the
fields
that
are
mentioned
in
this
in
this
slide
deck
right.
So
I'm
following
that.
B
Can
we
can
we
can
we
work
on
you
know?
Can
we
also
prioritize
things
like
validation,
since
we
are
trying
to
do
the
demo
as
a
development
milestone.
B
That's
right
sounds
good,
and
then
I
had
a
few
more
questions.
So
as
the
next
step,
so
you're
saying
we
already
have
bucket
class,
create
controller
in
place
right,
yes,
okay,
and
what
any
any
specific
actions
need
to
be
taken
for
bucket
class,
create.
C
Well
right
now
I
have
to
still
integrate
the
web
hook
so
that
we
do
web
hook,
validation
to
inject
the
default
class.
Things
like
that
and
yeah.
B
B
C
B
Okay,
so
so
that
that
might
not
work,
are
we
saying
that
we'll
always
have
a
default
buggy
class.
B
So
yeah
they
can
be
so
good
way
to
put
it
so
ben.
Was
there
you.
C
F
And
you
know
in
the
in
the
case
of
like
a
an
existing
pv
case,
it
might
just
bind
to
an
existing
pv
and
you
never
need
a
storage
class
that
that's
one
possibility
that
it's
less
interesting
from
this
perspective.
But
that's
one
thing
that
can
happen.
The
other
thing
that
can
happen
is
if
there
is
a
default
storage
class,
you
can
get
picked
up
and
assigned
automatically.
F
So
so,
like
the
the
way
you
designate,
you
can
actually
have
multiple
default
storage
classes
and,
like
that's
just
an
error,
but
like
it's
something
for
the
again
the
the
controller
to
resolve.
So
it's
nothing
wrong
with
not
specifying
a
storage
class
when
there
either
is,
are
zero
defaults
or
more
than
one
default,
both
of
which
let.
B
F
F
Bucket
class
is
created
now
I
think
that
the
key
would
be
like
the
controller
would
see
the
thing
get
created
and
it
would
say
I
don't
know
what
to
do
right
now.
I
might
generate.
B
G
D
C
F
There
was
recently
a
whole
fight
in
the
on
the
this.
I
don't
want
to
say
it's
a
fight.
There
was
a
heated
discussion
on
the
sig
storage
site
about
what
to
do
about.
You
know
the
fact
that
storage
classes
can
mutate
after
something
is
created
from
that
storage
class.
Like
do
you
use
the
version
of
the
storage
class
that
exists
when
you
created
it,
or
do
you
use
the
version
of
the
storage
class
that
exists
now?
What.
F
Well,
so
it's
immutable
right
technically,
yes,
but
you
can
always
delete
it
and
create
a
new
one
with
the
same
name.
G
Have
a
name
we
have
an
answer
to
that:
whatever
we
need,
the
bucket
class
is
only
at
bucket
creation,
time
or
say
brownfield.
You
know
sort
of
the
allowed
namespaces
time,
but
in
the
where
we
used
information
from
the
bucket
class,
those
fields
are
copied
to
the
bucket
instance
itself.
F
I
don't
think
you
sh,
I
mean
yeah.
I
just
I
think
you
need
to
you
know,
put
a
stake
in
the
ground
on
like
does.
Is
it
relevant
after
creation
or
not
because
if
it's
not,
then
on
the
brownfield
case,
I
don't
understand,
I
mean
I
I'm
sorry
I
was
on
vacation
last
week,
and
so
maybe
I
missed
like
an
important
meeting
about
what's
going
on
with
the
brownfield
case
but
like.
Why
would
we
consult
the
class
at
import
time
if,
if,
if
the
admin
already
has
to
create
the
bucket
object,.
F
B
F
Like
my
memory
of
how
the
import
works,
is
you
create
the
bucket
first
as
an
admin
with
all
of
the
relevant
details
that
would
have
come
from
the
store
or
the
bucket
class?
And
then
you
just
create
a
bucket
request
that
binds
to
it
and
off
you
go,
and
so
you
never
needed
a
bucket
class,
because
everything
was
already
on
the
object
right.
B
That
was
the
that
was
the
understanding.
However,
jeff
just
suggested
a
different
workflow
jeff.
Can
you
explain
it.
G
Yeah
yeah,
I
mean,
like
I
said:
we've
gone
back
and
forth
a
number
of
times.
What's
the
final
conclusion,
I
don't
know
if
it's
final,
the
cap
says
the
bucket
class
is
required
for
both
greenfield
and
brownfield
with
the
reason
being
the
allowed
namespaces
is
the
only
other
than
our
back
is.
Is
an
additional
control
point
that
admin
has.
However,.
F
G
F
D
F
It's
it's
an
intentional
process.
It's
not
automatic,
then
someone's
going
to
have
to
decide
what
to
put
in
that
field
and
they
can
go
look
at
a
bucket
class
if
they
want
to
or
they
can
just
populate
with,
with
whatever
they
want,
and
so
the
I
in
either
case
the
bucket
class
will
get
ignored
because
you
put
something
in
the
field
when
you
created
the
object.
G
B
I
agree
jeff,
I
think
I
think
you
know
this
is
this
has
been
here
from
a
previous
version
of
the
cap
and
I
think
can
you
you
need
to
remove
that
line.
A
By
andrew,
this
is
wrong.
The
class
name
comment,
or
you
guys
remember,
but
I
also
agree
with
ben-
seems
to
be
yeah
not
really
needed
for
from
yeah.
G
You
know,
there's
it's
a
weak
argument,
I
think,
but
another
argument
would
just
be
consistency
in
in
brs,
and
one
of
our
more
minor
goals
was
that
there
was
to
not
have
a
lot
of
differences
between
brownfield
and
greenfield.
So
why
do
we
need
a
br
for
braun
field.
F
C
F
Yeah
agreed,
but,
but
I
I
think
this
is
the
essential
difference
between
the
greenfield
and
the
brownfield
is,
is
to
create
a
greenfield
one.
You
need
a
class
and
to
create
a
brownfield
one.
You
don't
and
that's
why
they're
different,
so
the
consistency
in
general
is
important,
but
this
is
one
of
the
important
differences.
F
F
B
Makes
sense,
I
think,
that's
what
I've
seen
too.
I
think
it
can
be
ignored
if
it's
brown
field
yeah,
it
need
not
be
an
error,
condition.
B
Yeah:
okay,
we
have
only
five
minutes
left,
okay,
so
extremely
the
last
one
is
bucket
access
request.
C
I
did
the
same
thing
for
the
bucket
access
request.
I
have
a
controller
which
acts
on
the
bucket
access
request
and
then
copies
the
fields
from
the
bucket
access
request
and
the
bucket
access
class
into
the
uncreates,
the
bucket
diagram,
okay,
so.
C
The
validation
is
again
the
same
I
mean
if
I
don't
find
the
bucket
access
class,
I
throw
an
error
back
written
an
error
but
yeah
that
still
makes
sense,
because
that
way
we
can
retry
right.
Okay,
so.
C
B
B
Okay,
now
that's
fine,
I'm
just
trying
to
track
the
progress
and
yeah
we'll
we'll
we'll
go
step
by
step
again.
This
piece
of
code.
B
Yeah
before
we
can
actually
also,
you
know,
it'll
help
when
we
get
the
official
kubernetes.
D
B
Because
there
we
will
be
able
to
get
reviews
from
a
lot
of
people,
yeah.
C
One
more
thing
here
is
that
we
could
not
use
storage,
dot,
case.io
api
group,
because
there
is
already
one
with
v1
alpha
one
in
the
main
branch.
So
there
is
some
fcd
is
not
happy
with
it.
So
we
fall
back
to
object,
storage,.
B
Yeah
yeah,
we
can
explain
that
I
want
to
quickly
finish
the
tasks.
B
Okay,
so
provisional
control
is
rob
on
the
call
yep
okay,
so
I
can
I
can
fill
in
for
what
last
I
checked
with
rob.
We
don't
have
a
cli
in
place.
Yet,
however,
we
have
the
basic
bucket
create
controller
in
place
without
any
logic
in
there
and
for
writing.
The
sidecar,
a
new
person
has
joined
with
you
know,
start
contributing
to
the
development
effort
sergeant
jaiswal,
I
believe,
he's
on
the
call
today,
yeah
hey
did
you?
B
Sweet
okay,
looking
forward
to
that,
and
then
chris
are
you
on
the
call?
B
Oh
yeah
hey
chris,
do
you
want
to
give
us
an
update
on
where
you
are
with
the
cs?
Adapter.
H
No
major
updates
on
the
csi
adapter
I've
just
been
adding
on
to
the
existing
code
that
we
discussed
last
week,
and
we
had
some
further
discussion
over
the
week
regarding
some
specific
implementation.
Details.
H
Such
as,
like
you
know
where
the
file
path
will
be
mounted,
and
so
I've
just
been
going
through
and
adding
support
for
that
everywhere.
B
Okay,
got
it
okay,
so
are
we
at
the
stage
where
we
create
the
ephemeral
volume.
B
Yeah
so
because
it's
copied
from
the
previous
slide,
we
still
have
this
entry
here.
Okay
sounds
good,
so
I
have
a
good
idea
of
what
the
progress
is.
Is
there
anything
else
anybody
wanted
to.
C
B
Yeah,
so
the
api
okay,
so
we
still
have
a
minute
good.
So
the
api
group
discussion
is
pretty
straightforward.
So
in
order
to
so
a
few
weeks
ago,
we
decided
that
the
api
group
that
all
the
bucket
objects
bucket
star
objects
should
reside
in
was
storage,
dot,
k
destroyed,
slash,
we
don't
have
one
to
start
with,
and
one
thing
we
noticed
is
in
order
to
create.
B
B
So
for
now,
just
until
you
know
either
we
have
to
get
the
bucket
objects
merged
into
the
api
repository,
which
requires
an
api
review
or
until
then
we
have
fallen
back
to
object,
storage.ktest.io,
which
we
can
utilize
for
now
and
when
we
do
get
things
merged
into
the
official
api
repo,
we
can
change
it.
B
Okay,
all
right
all
right
that
that's
all
from
my
side.
We
have
we,
we
can
discuss
anything
else.
If
you
have.
B
Oh
yeah,
so
yes
right,
we've
added
some
some
boilerplate
work
right
like
issue
templates
and
something
else.
So
the.
C
Ti
and
as
well
as
chris
added
those
templates
for
the
issues
and
pr's.
Yes,.
C
B
Thanks
all
right,
thank
you.
Everyone,
if
you
have
any
other
questions,
please
reach
out
to
us
on
the
slack
channel,
sig,
storage,
cozy
and
please
take
just
few
minutes
out
of
your
day
and
and
review
the
cap.
We
want
to
get
it
merged
this
week.
G
Go
ahead
jeff
last
week,
saad
reviewed
the
cap
and
had
about
six
comments
and
feedback,
and
we've
addressed
those
I'd
like
to
see.
If
you
agree
with
us
on
on
how
we
responded.
B
And,
and
also
so
one
thing
yeah,
we
we
will
update
the
cap,
and
you
know
talk
about
the
bucket
class
based
on
the
discussion
today,
where
it's
not
needed
for
brownfield.