►
Description
Kubernetes Storage Special-Interest-Group (SIG) Object Bucket API Standup Meeting - 25 January 2021
Meeting Notes/Agenda: -
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
A
Can
you
all
hear
me?
Hey,
yes,
hey,
okay!
So
again,
let's
quickly
go
over
where
we
are,
we
made
a.
We
made
good
progress
during
the
last
three
days,
so
rob
had
to
step
away
for
a
few
days
due
to
a
personal
emergency.
He's
fine
everything's,
okay,
it's
just
he.
He.
A
Only
contributing
probably
mid
this
week
or
late
this
week,
like
at
the
end
of
this
week,
I
would
say:
that's
what
he
told
me
he'd
be
back
sooner,
but
you
know
you
never
know.
So,
just
that
update.
We
have
a
spec
pr
waiting
on
some
changes,
responses
to
comments.
I
think
we
addressed
them
last
week
and
you
know
that
needs
to
be
done
and,
and
I'm
the
one
that
is,
that
is
working
on
that
right
now.
A
Okay,
so
that
that
was
a
quick
update
about
what's
happening
with
that.
Now,
let's,
let's,
let's
actually
quickly
go
with
the
milestones
and-
and
I
want
to
get
an
update
from
everyone
on
on
what
they're
working
on
so
again,
the
current
milestone
is
demo
and
api
review
the
deadline
for
api
reviews
coming
up
fast
or
for
submitting
the
api
reviews
coming
up
fast.
A
We
have
just
one
pull
request
in
the
spec
repository
that
needs
to
be
merged
before
we
go
up
to
the
api
reviewers
and
ask
them
to
start
looking
at
our
proposal,
and
I
believe
we
need
to
update
the
cap
as
well
once
the
spec
is
updated.
Is
that
right,
shane.
A
Yeah
no
way
so
so
we
need
to
update
the
cap
as
well
right
before
they
appear.
B
B
Also,
you
need
a
production
revenues
review
that
should
be
easy.
Just,
I
think,
there's
only
like
a
few
questions
regarding
the
future
gate
like
how
you
enable
it
things
like
that,
so
it
should
be
easy.
B
Yeah
that's
correct,
so
there
is
a
section
about
production
revenues.
I
think
you
should
just
fill
out
the.
I
think
there's
only
one
section
that
you
need
to
figure
out,
which
is
about
the
feature:
how
to
enable
the
feature
things
like
that
got.
A
B
A
I
I
think
we've
made
quite
a
few
changes
between
when
this
was
merged
and
right
now
I
think
we
have
to
update
the
apis
and
and
also
descriptions
based
on
all
the
discussions.
We've
had
right
right,
jeff
if
you're
there
or
sweeney.
D
That's
true
yeah,
and
would
you
know
what
those
changes
are?
We
have
to
wait
for
the
pr
on
the
spec
repo
and
once
that
is
done,
it's
a
lot
easier
for
us
to
update
the
cap.
A
So
so
the
the
the
key
thing
that
I
still
I
still
need
to
finish
is
auto,
generating
the
kohli.proto
file
from
the
spec.
I'm.
What.
B
A
Right,
I
see
what
you
mean.
Okay,
I
mean
it's
not
too
much
work,
it's
just
something
that
needs
to
be
done.
Okay,.
A
Yeah,
it
shouldn't
take
long,
probably
the
end
of
the
day.
Today
I
should,
I
should
have
all
of
it
put
together.
Yeah
yeah
one
skip
is
merged,
so
sweeney.
Do
we
also
have
to
make
changes
to
the
api
api,
repo
api.
D
Repos
changes
we
yeah,
we
did
that
and
the
pr
is
out.
So
we
don't
have
a
issue
with
that,
but
I
have
another
pr
in
the
spec
repo
that
jeff
had
made
more
comments
over
the
weekend.
I
fixed
them,
so
I
pinching
to
take
a
final
look
at
it
because
that's
we
decided
we
are
going
to
extract
the
the
proto
dot
protocol
file
from
this
right.
D
If
that
I
already
tested
it,
I
haven't
issued
a
pr
yet,
but
it
will
extract
the
data
from
it
appropriately,
but
I
might
go
through
this
in
couple
of
iterations
right
now.
The
entire
dot
prototype
is
one
section
in
the
document.
The
basis
I
did
is
not.
D
Yeah
we
can
do
that,
but
main
focus
is
right
now
having
these
messages
clearly
marked,
so
that
they
will
be
extracted
out
by
the
makefile
to
build
the
product.
A
Yeah
so
the
way
I
was
thinking
about
it
was
we
should
just
have
these
three
backticks,
followed
by
say,
protobuf.
D
Yeah,
that's
it!
How
do
you
find
the
end
by
the
three
games.
A
A
B
D
That
they
had
each
of
the
messages
separated
out
instead
of
a
single
blob
like
this,
that's
what
I
was
thinking
too.
I
already
have
them
some
of
them,
but
not
all
of
them.
D
A
Okay,
I
think
this
is
fine
to
start
with,
like
you're
saying
okay.
So
what
about
this?
Do
you
have
a
make
generator
or
something
to
yeah,
then
that
we
can
directly
use
it
from
csi
and
it
works
right.
D
Okay,
so
does
this?
Is
it
like
make
generate
or
what
starts
to
change
it?
Will
it
will
extract
this
into
a
dot
proto
file,
and
then
it
will
generate
the
proto
in
group.
A
Are
these?
Are
they
two
steps
or
just
one?
It's
all
done
in
the
make
when
you
do
a
one
step?
Okay,
that's
fine,
all
right!
A
So,
okay,
so
I
I
made
a
pull
request
yesterday
on
the
readme
file,
because
I'll
be
honest
with
you,
the
readme
does
not
make
much
sense
right
now,
like
this
part,
it's
using
html,
that's
the
one
part
I
want
to
get
rid
of,
and
also
this
doesn't
mean
anything
right
now,
so
this
also
needs
to
be
updated
as
people
come
and
look
at
look
at
this
project,
it's
hard
for
them
to
tell
what
it
does
so
or
how
to
get
started
with
this.
A
So
this
pr
includes
all
of
rob's
changes,
so
it
should
not
be
merged
until
drop
speed
is
merged,
but
but
here's
the
here's,
the
new,
updated
readme
file.
A
You
know,
I
think
we
should
explain
what
this
is
and
then
also
explain
why
this
is
needed
and
then
here
how
to
get
started
or
how
to
make
changes
to
this
repository.
I've
created
it
as
a
separate
step
make
generate.
I
see
yeah.
D
A
I
see
if
we
can
do
this.
That
would
be.
That
would
be
good,
so
you
can
do
the
do
it
in
two
steps
and
make
all
should
just
do
everything
yeah.
It's.
D
A
Stopping
us
from
doing
it
this
way,
how
much
time
do
you
think
it
would
take
it's
it
shouldn't
be
hard
at
all.
Okay,
could
you
could
you
take
care
of
this
journey?
That'll,
be
great
awesome,
okay,
all
right,
so
so
that
that
means
yeah
I'll
be
done
even
sooner
thanks
to
srini
appreciate
it
all
right.
So
so,
once
we're
done
with
that,
if
I
understand
correctly,
the
spec
is
sorry.
The
api
app
was
in
good
shape.
What
we
need
after
that
is
changes
the
cap
correct,
correct,
yeah,
all
right!
A
So
sorry,
jeff
is
not
here,
so
I'll
follow
up
with
jeff
on
where
he
is
with
the
cap.
If,
if
jeff
is
busy
so
actually
hold
on
so
sh
shing,
the
only
change
we
have
to
do
is
make
it
implementable,
correct
or
production
ready.
D
B
B
Yeah
same
but
I'm
saying
like
like
for
other
features,
we
also
have
the
envelopes
right,
like
other
six
storage
features
they're
all
like
out
of
three,
but
you
just
still
need
to
answer
those
questions
yeah.
So
it
should
be
easy.
Just
only
the
first
section
that
is
required
for
alpha
cap.
A
A
And
also
the
spec
at
the
end,
so
I
think
you
know
we'll
just
yeah
we'll
just
make
updates
to
the
spec
as
it
is.
I
mean
in
the
sense
we'll
just
copy
over
from
the
pr
that's
there
and
you
know
if
the
pr
changes
will
update
here
also,
but
but
I
think
we
should
get
started
on
the
kepir
sooner
rather
than
later,
like
you're
saying
yeah.
D
A
Yeah,
okay,
so
yeah
I
mean
the
logistics
are
fine.
I
think
it's
okay,
we
can.
We
can
handle
this
all
right,
so,
okay.
So
there
is
another
question
about
spec
that
I
want
to
bring
up
it's
a
little
more
technical
and
implementation
oriented.
A
A
A
There
are
a
few
concerns
that
we
ran
into
with
this
with
this
idea,
that
is
say
in
the
api
repo,
if
you
could
just
import
this
one
structure,
instead
of
having
to
redefine,
say
the
protocol
structure
for
s3
here
and
in
there
in
the
api
repo.
If
you
could
just
import
it,
then
you
know
we'd
be
following
programming
best
practices,
which
is
don't
copy
code.
However,
we
did
run
into
one
issue:
that
is
the
crds.
A
The
cid
generation
fails
because
of
these
non-imported
fields.
So
does
anyone
here
know
if
there
is
a
way
to
include
what
above
generated
data
structures
inside
kubernetes
api.
A
E
F
Yeah,
it's
not
jrpc,
so
so
do
we
really
want
them
to
be
this?
Well,
they
always
evolve
at
the
same
pace.
Yeah.
That's
a
good
question.
E
A
Yeah
as
long
as
they're
backwards
compatible
yeah,
we
should
be
able
to
work
with
those
changes
now.
My
the
reason
for
considering
this
is
it's
better
not
to
copy
code
anywhere
or
try
to
keep
things
in
sync
manually
by
hand.
Rather,
if
you
could
just
import
it,
there
are,
there
are
less
chances
of
incongruence
between
the
two
data
structures
between
the
two
types.
E
B
E
A
Using
protobuf
will
will
make
sure
that
it
actually
fails,
because
if
you,
if
you
were
to
remove
a
field,
then
product
of
parser
will
fail.
If
you
move
it
on,
I
see
what
you
mean.
If
you
move
it
on
both
sides,
it's
not
going
to
be
the
problem.
Okay,
yeah,
you're!
Right
again,
it's
it's
something
to
consider
it's!
I
look
at
it
as
it's
better,
not
to
duplicate
it.
That
way
if
possible.
A
However,
like
you
said
feed,
I
mean
by
hand,
we
can
obviously
write
a
crd
based
on
the
structure
and
because
these
fields
are
not
imported,
it
works
just
fine,
but
only
if
you
do
it
by
hand
and
if
you
were
to
build
a
cid
by
hand,
I
mean
you're,
not
really
reducing
the
risk
of
making
manual
mistakes
so
yeah.
I
we
later
on
found
that
it's
it's
as
of
right.
Now,
it's
better,
probably
not
to
go
down
this
path
and
just
have
two
different
definitions.
A
E
A
Okay
moving
forward,
so
so
we
have
we,
we
discussed
four
important
requirements
for
doing
the
demo.
When
we
said
we'd
do
the
demo.
We
said
we.
We
said
that
when
the
demo
is
done,
the
the
code
will
actually
be
usable
and
it
will
be
a
high
quality
code
that
is
well
tested
and-
and
you
know,
it's
reviewed-
not
just
not
just
a
smoke
and
mirrors
demo.
To
that
end,
I
wanna
again
quickly
give
an
update
on
where
we
are
so
in
terms
of
development.
A
We
said
we
would
have
the
three
simple
features
of
creating
a
bucket,
creating
access
for
that
bucket
and
pro
provisioning
that
bucket
for
a
pod.
A
So
we
I'm
happy
to
say
we
have
these
three
features
implemented
and
we've
tested
it.
It
works,
it
works
with
our
sample
driver
and
we
did
have
to
go
ahead
and
make
some
updates
to
the
grpc
expect
to
get
this
working.
A
A
A
We
had
only
one
generic
map
string
string
in
the
in
the
create
bucket
request
definition,
and
so
the
only
way
to
move
ahead
with
that
such
a
definition
was
to
was
to
combine
these
two
maps
into
one
over
here.
A
All
the
benefits
that
we
were
going
to
get
from
having
protocol
be
a
strictly
type,
strongly
type
structure
would
basically
get
thrown
out
the
door
when
bucket
context
is
a
generic
map
that
combines
two
different
maps,
two
different
structures.
A
So
the
question
that
we
brought
up
was:
should
we
make
this
part
into
two
different
fields
where
we
have
protocol
and
parameters
and
protocol
here?
Also
in
the
grpc
spec
would
be
strongly
typed
just
like
in
the
api,
and
so
the
unanimous
answer
is
yes.
A
It
is
better
to
have
protocol,
be
strongly
typed
in
the
grpc
api
as
well,
and
to
have
a
separate
generic
map
string
string
for
parameters,
just
as
it
is
in
the
bucket
class.
So
it
reflects
you
know
evenly,
and
so
this
is
the
change
that
led
us
to
having
a
protocol
structure
defined
in
the
grpc
spec
and
the
protocol
structure
defined
separately
in
the
api
repo,
as
well
nicholas.
Does
that?
Does
that
give
you
the
reasons
why
we
went
that
way.
A
Okay,
I'm
gonna
move
forward,
so
so
in
terms
of
development,
we
are
in
pretty
good
shape.
Now
that
this
change
has
been
done.
We
are
waiting
for
this
to
be
merged
before
updating
the
controller
sidecar
and
the
csr
adapter
to
you
know,
pull
in
these
changes
and
we
believe
it
will
be
a
very
quick
change.
So
that's
in
terms
of
development.
A
The
next
thing
we
focused
on
was
deployment.
We
have
that
under
control
we
in
order
to
set
it
up
in
order
to
set
up
the
whole
system.
You
just
need
to
run
four
cube,
ctl
commands,
so
you
just
have
to
do.
Cubesat
create
dash
k,
k
for
customize
and
the
name
of
each
of
the
repos,
so
the
api
repo
would
register
the
crds
controller
would
install
the
controller.
Sidecar
would
install
the
sidecar.
The
css
adapter
would
install
the
node
adapter.
A
This
has
been
tested
multiple
times
it's
in
good
shape
the
next.
The
next
thing
is
testing.
We
have
unit
tests
and
ci
in
place
in
terms
of
end-to-end
tests.
I
was
saying
this
last
week,
I
think,
before
the
demo
waiting
for
the
e2e
tests
to
be
merged
requires
allocation
of
resources
for
cosy
specifically,
and
we
really
don't
have
control
over
that
timeline.
A
So
the
best
way
we
can
we
can
address
the
that
is
to
manually
do
the
et
testing
by
the
demo
time.
So
that's
the
that's.
A
The
one
thing
I
brought
up
last
week
about
testing
the
final
thing
is
documentation,
so
in
the
pr
that
I
showed
you
earlier
made
to
the
spec
repo
we've
also
got
prs
for
it
or
in
peers
on
the
works
for
the
other
repos
and
those
are
the
getting
started
guides
just
quick
introduction
to
what
this
pr,
what
this
repository
is
for
and
how
to
get
started
to
build
it
or
make
changes.
A
Things
like
that,
and
the
spec
guide
is
the
pr
that
srini
was
talking
about
where
he's
going
to
auto
generate
the
protophile
from
the
spectromd.
So
again,
this
is
very
close
to
being
done.
The
prs
are
in
the
final
stages
of
review
and
approval,
but
before
the
demo
we
want
to
make
sure
that
all
of
these
are
green.
F
One,
if
you
could
go
back
two
slides,
I
believe:
okay,
two
slides,
okay.
Yes,
I
just
this
could
cause
some
some
confusion,
because
here
we
are
saying,
install
the
sidecar
as
the
the
third
step.
Where
and
I
I
don't
think
anyone
would.
If
you
you
deploy
cozy,
you
will
never
really
install
the
sidecar
right
right,
so
I
should
be
able
to.
D
A
Yeah,
I
think
we
should
still
have
something
that
I
I
completely
agree.
First
of
all,
yes
you're
right,
so,
first
of
all,
when
I
say
sidecar
I
mean
the
provisioner
costs
the
sample
forever,
and
that
needs
to
be
clarified.
Yes,
right
now,
the
way
it
is
right,
the
way
it
is
it
it
helps
you.
You
know
it
serves
as
a
guide
for
someone
who
is
reading
the
customized
files
to
figure
out
how
to
integrate
the
provisioner
and
the
sample
driver.
A
We've
just
used
the
catch
all
term
for
calling
it
sidecar
when
we're
running
both-
and
I
agree
we
need
to-
we
need
to
make
some
changes
there
in
terms
of
how
yeah,
like
you
said,
the
architecture
and
also
even
in
the
deployment
guide,
I
think
we
can.
We
can
have
it
be
two
separate
steps.
Maybe
I'll
explain
it
better.
A
Agree,
perfect,
okay,
so
all
right,
so
it's
it's
11!
30.,
we've
run
out
of
time.
I
think
ben
says
he
also
has
to
go
right
now.
Do
you
have
any
questions?
A
No,
all
right
so
by
thursday.
Look
forward
to
more
updates
or
more
progress
on
on
these
four
steps.
Ben
was
asking
last
week
if
there
is
a
sample
driver
that
he
can
take
a
look
at
and
start
playing
with.
The
answer
was
yes,
I'm
gonna
show
it
again
for
people
who
are
not
aware,
so
we
have
a
working
sample
driver
under
a
provisional
site
card.
Cmd,
slash
sample,
driver.
A
And
yeah
it's
about
143
lines
and
it's
a
very
basic
one
that
that
creates
a
minio
bucket
and
provisions
access
for
it,
revokes
access
and
deletes
it's
not
sophisticated.
It's
very
basic.
It
doesn't
do
you
know
every
kind
of
all
kinds
of
use
cases,
but
but
it's
a
good
start
for
people
who
want
to
start
writing
their
own
sample.
A
Drivers
all
right,
so
that's
it
from
me
if
you
have
questions
again
reach
out
to
me
or
anyone
on
the
slack
channel
six
doors,
cozy.