►
Description
Kubernetes Storage Special-Interest-Group (SIG) Object Bucket API Review Meeting - 10 December 2020
Meeting Notes/Agenda: -
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
A
Key
changes,
starting
with
our
project
management
tools,
all
thanks
to
nicholas.
I
want
to
single
him
out
and
thank
him
because
nicholas
brought
up
last
week
or
maybe
a
week
before
that
saying
that
it
has
been
hard
for
him
to
get
on
board
on
the
project
and
he
wasn't
sure
where
to
get
started.
I'm
glad
he
brought
it
up
and
said
it
that
way,
because
it
it
immediately
me
that
that
needs
to
be
fixed.
A
We
were
managing
our
projects
using
just
our
own
tools
and
weren't,
really
doing
it
on
github,
which
is
generally
idiomatic
with
open
source
projects.
A
So
under
kubernetes
6,
I
believe,
a
few
months
ago,
container
object
storage
project
was
created,
so
I'm
just
going
to
open
that
page
and
this
tuesday
we
all
met,
including
nicholas,
and
we
went
over
how
we're
going
to
manage
this
project
using
this
github
tool
data
projects
tool.
So
what
we've
defined
is
a
milestone.
A
We've
called
it
demo,
one
as
you
can
see
in
many
of
the
issues
here
and
all
the
issues
that
are
to
do's
they're,
obviously
going
to
to
do
in
progress
is
when
someone
has
started
working
on
it
in
review
is
when
they've
made
the
pr
and
it's
being
reviewed,
and
it's
done
it's
done
for
anyone
that
wants
to
get
onboarded.
A
The
easiest
direction
we
can
give
them
is
to
go
to
this
page
and
filter
by
good
first
issue,
and
they
can
get
started
if
someone
who
can
is
there
who
is
interested
in
writing,
documentation
they
search
by
that.
A
Okay,
how
do
I
remove
the
filters
there
yeah
and
we're
going
to
follow
this
labeling
scheme
in
order
to
allow
this
easy
filtering
for
people
to
discover
what
is
going
on
and
and
how
much
progress
has
been
made.
So
for
that
I
can
say
milestone.
A
A
One,
oh
like
that,
not
sure
what
the
format
is
there,
an
easier
way
would
be
to
just
click
on
that
yeah.
Oh,
I
got
that
format.
Okay,
so
you
can
also
filter
by
milestones
and
see
how
much
progress
we've
made.
A
The
good
thing
about
this
structure
that
we
brought
up-
we
we
did
this
live
on
a
video
call
on
tuesday
on
wednesday
morning
there
were,
there
was
a
new
contributor,
including
nicholas.
This
person
is
from
bangalore,
india.
We
haven't
talked
to
them
yet
other
than
through
github
issues.
A
It
was
really
nice
to
see
that
this,
this
sort
of
management
is
helping
you,
even
if
it's
just
one
person,
it's
definitely
in
the
right
direction
of
getting
more
people
on
board
to
the
project,
and
once
the
the
project
management
was
clearer,
we
could
also
start
iterating
faster.
The
division
of
tasks
was
clearer
and
that
also
helped
us
move
faster
and
compared
to
the
last
few
weeks
this
week
alone.
A
These
are
all
pr's
from
this
week
alone
and
the
activity
from
this
week
alone
so
generally
just
a
pr
could
be
even
a
one-line
change,
so
it
doesn't
mean
anything
but
in
open
source
projects,
pr's
are
generally
the
bottleneck,
because
we're
not
not
not
all
of
us
is
doing
this
as
a
full-time,
gig
and
sometimes
reviews
take
longer.
Sometimes,
responses
to
reviews
take
longer
so
prs
being
the
biggest
bottleneck
are
still
a
good
indicator
of
progress
in
open
source
projects.
A
So
here
here
is
how
we've
been
doing
last
week
and
what
we
want
to
get
done
before
18th,
which
is
the
end
of
next
week,
is
we
want
to
ramp
up
the
documentation
a
little
bit
because,
as
it
is
today,
the
documentation
is
outdated.
A
Now,
first
of
all,
I've
added
issues
for
it.
If
you
specifically
look
at
csi
adapter.
A
Yeah,
this
documentation
can
be
changed.
This
is
pointing
to
the
old
slack
sorry
six
stories
slack,
which
is
not
six
stories
cozy.
So
these
changes
need
to
be
undertaken,
and
hopefully
before
the
end
of
the
year,
so
that
when
people
come
in
fresh
next
year,
it's
just
they're
ready
for
them
to
consume
all
right.
So
going
back
here.
A
A
B
Yes,
for
all
the
reapers
now
voice
started
only
midway.
Oh
sorry,
yeah
there
is
some
lag
in
my
zoom,
so
anyways
as
far
as
the
pre-submit
jobs
are
concerned,
controller
has
it
now
spec
and
api.
Has
it
for
the
sidecar
and
see
I
said
after
I
have
submitted
a
pr
and
that's
thanks
to
xing.
She
reviewed
it
and
we
just
need
one
more
reviewer
to
merge
that
for
the
first.
A
B
No
for
the
csi
adapter
and
the
procedure:
okay-
okay,
that's
the
last
pr
for
the
pre-submit.
For
this
pack.
There
is
pro
data
sh
that
I
I
changed,
but
it's
work
in
progress
because
I'm
waiting
for
some
of
the
code
to
be
checked
in
so
that
I'll
have
a
better
understanding
on
how
to
load
up
various
artifacts
to
do
testing
inside
the
client
cluster
that
pro
spins
up.
B
So
it
will
xing
already
reviewed
it
except
few
changes
that
I
need
to
make
to
it,
so
that
it
is
a
little
bit
more
parametrized
so
that
we
can
test
various
branches.
What.
C
B
B
Absolutely
locally
we
can,
we
can
still
build
the
containers,
we
can
push
the
containers
to
our.
We
have
created
a
temporary
repo
in
queer
and
we
are
using
that.
I
already
pushed
an
image
for
controllers,
so
it's
a
workaround,
but
you
know
for
newcomers
coming
into
the
project.
They
need
to
understand
that,
rather
than
that,
it
would
be
nice
that
if
we
have
okay
integrated.
A
B
I
was
about
to
say
I
mean
you
showed
that
documentation
page
it.
It
has
to
explain
how
what
the
that
component
does
and
how
to
build
the
component
and
push
the
component
for
testing
and
how
to
deploy
it,
because
we
already
have
some
deployment,
scripts
and
whatnot
right
in
each
of
the
repos
and
that
that
are
in
progress
right
now
I
mean
they
are
they're
part
of
the
old
repo.
They
are
being
moved
to
the
new
repo.
So
once
we
have
that
it
is
a
lot
easier
to
deploy
individual
components.
B
Okay,
I
would
I
I
don't
know
I
mean,
maybe
you
can
comment
on
this,
but
generally
for
for
for
building
and
testing
as
part
of
the
post
submit.
We
don't
need
to
do
anything
it
just
push
to
gcr
and
get
it
will
do
the
work
done
apart
from
the
training.
D
Is
actually
I
mean,
I
think,
the
image
part
it's
not
completely
automatic
yet,
but
I
think
you
are
not
there
yet
so
I
mean
you
first,
you
need
to
you
need
to
cut
a
release
before
we
can
actually
have
those
images.
D
Once
you
are
there,
we
can.
We
can
go
through
that,
but
right
now
I
think
you
still
have
other
things
that
you
need
to
fix.
First
right.
D
You
have
to
have
like
a
release
cut
actually
so
right
now
we
don't
have
to
have
a
way
to
to
publish
like
in
the
past.
We
have
those
like
canary
images
like
whatever
that
is
in
the
master.
We
don't
have
that
anymore.
So
it's
a
little
bit.
It's
a
kind
of
annoying,
actually
yeah.
So
the
only
thing
that
will
be
there
will
be
a
formal
release,
so
it
has
to
have
a
like
a
tag,
something.
A
D
E
I
o
is
free,
but
sorry
what
is
three
key
dot
io,
the
one
that
oh.
F
D
D
E
D
Or
not
a
problem,
I
just,
I
think
they
want
to
try
something
right.
I
mean
the
the
problem.
Is
you
only
release
the
image
you
can
go
there
in
the
past?
Remember
we
have
the
canary
images
published.
D
D
E
D
Oh,
that's
the
thing
yeah
yeah.
I
think
I
think
they
might.
I
think
they
were
saying
that
there
was
a
way
to
put
that
in
like
the
staging
area,
but
then
I
think
you
have
to
constantly
pulling
it
because
otherwise
you'll
be
removed.
D
E
Wow
so
so
I
was
gonna
say
for
for
just
doing
personal
testing,
like
it's
very
easy
to
get
your
own
repo
on
key
dot,
io.
D
But
when
you're
testing,
whatever
that
is
in
the
master,
then
that's
just
to
use
whatever
it
is
build
from
the
master
from
your
pr.
For
that
you
don't
really
need
an
image
unless,
if
you're
testing
some
older
build
right,
so
I
don't
think
you
guys
are
there
yet
like
we
are
testing.
For
example,
other
side
cars
were
testing
with
like
1.17
minor
18,
so
you
don't
have
data
yet
right.
So
you,
if
you're
just
testing
this
like
one
ripple.
B
A
B
D
Basically,
like
let's
say,
if
you
are
testing
it
in
the
controller
repo,
then
whatever
pr
in
the
controller
repo
that
will
be
built
right,
you
build
from
that
ripple
so
that
that's
like
the
fresh
one,
the
new
one
everything
else
will
be
from
image.
Basically,
so
that's
yeah,
that's
how
you
can
test
it.
I
think.
A
Okay,
yeah
we'll
we'll
do
that.
I
think
that
sounds
good,
especially
after
the
demo.
Until
then,
we'll
just
use
key
or
io.
A
A
I'll
show
you
and
also
you
know
it
took
me
a
few
tries
before
I
started
getting
it
right,
it's
container
object,
storage,
interface,
controller
and
then
it
gets
longer.
So
the
sidecar
it's
provisional
cycle
and.
B
If,
if
anyone
knows
that's
fine,
but
if
not,
I
can
think
some
of
these
really
folks
and
see
if
we
can
rename
the
repos,
but.
A
Yeah,
that's
what
I'm
bringing
up
actually
so
there's
that
the
second
thing
I
have
is
so,
if
you
look
at
csi,
cni,
container
storage
interface,
container
network
interface,
they're
all
specific
to
the
container
itself,
they're
an
integration
with
the
container
so
regardless
of
who
the
container
provider
is
as
long
as
they
satisfy
oci
another
standard
open,
container
interface,
they
they
can
integrate
with
csi
and
cni
object.
Storage
is
a
little
different,
we're
not
really
integrating
with
the
container
we're
integrating
with
the
orchestrator
here.
A
A
A
I'm
open
to
opinions,
I'm
open
to
ideas.
This
is
a
opinion
kind
of
discussion,
but
one
thing
is
clear:
the
word
container
here
is:
can
can
even
be
little.
It's
not
idiomatic
with
how
the
rest
of
the
industry
is
naming
standards
and,
at
the
same
time
it's
not
completely
accurate.
A
E
A
A
I
mean
you
could
say
the
same
about
csi
or
cni.
There
are
a
lot
of
abbreviations.
I
mean
there
are
a
lot
of
things
called
csi
before
and
kubernetes,
because
it's
the
new
csi
it
took
over
osa
actually
had
two
different
definitions.
F
G
You
call
it
cozy
yeah
yeah,
no,
that
wasn't
it
yeah.
I
and
I
I
like
that
idea.
Actually
it's
a
it's
more
actor
descriptive.
It
addresses
ben's
comment,
yeah,
it's
more
specific
yeah.
G
And
plus
one
what
sid
just
said,
we
really
aren't
trying
like
csi
at
the
very
beginning,
wanted
to
be
kubernetes,
agnostic
and
and
we're
not
trying
to
do
that
really.
E
E
F
So
I
don't
think
it's
necessarily
a
misnomer
to
call
it
container
object,
storage,
interface,
it
you
know
they.
You
know
other
orchestrators
to
the
extent
that
they're
still
viable
could
you
know,
could
implement
this
interface,
even
though
you
know
we're
not.
We
don't
have
the
requirements
to
hook
into
o6
and
file
system
mounts
and
things
like
that.
We
have
less
requirements
in
the
container,
but
other
orchestrators.
A
A
E
So
so
so
the
way
the
way
I'm
looking
at
cozy
is,
if
we're
successful
and
two
years
down
the
road
we're
out
of
beta
and
it's
ga
and
there's
been,
you
know
three
or
four
different
protocols
or
sorry
yeah.
A
E
Yeah,
if
you
write
your
application
to
sit
on
top
of
the
cozy
grpc,
then
you
can
swap
in
any
different
any
object:
storage
provider
and
it
just
works,
and
so
it
becomes
this
common
way
to
sort
of
ask
for
buckets
to
receive
buckets
to
ask
for
access
to
buckets
to
get
the
the
protocol
information.
All
the
grpc
is
responsible
for
all
of
that
level
of
stuff,
and
people
outside
of
kubernetes
could
in
principle
be
interested
in
that,
even
if
they're
not
container
orchestrators,
they
could.
A
E
E
Yeah
I
I
buy
that
I
gotta
apologize,
that
I
have
a
conflict
now
and
I
gotta
drop.
But
oh.
A
A
I
think
adding
more
words
sometimes
does
is
like
it
doesn't
help
us
exactly
how
we
want
it
to
in
order
to
proliferate
this.
Keeping
it
osi
is
is
good.
Enough
is
good,
especially
given
that
the
other
osi
is
an
old
standard
and
it's
unlikely
to
come
up
in
the
same
context
and-
and
the
other
reason
is
even
if
you
came
up
with
ospi
or
posi,
or
something
like
that,
you
still
run
the
risk
of.
A
F
F
That
acronym
in
particular,
is
difficult,
like
I
think
that
the
conflict
isn't,
you
know,
a
trivial
concern
is
what
I'm
saying
and,
and
it's
there's
a
big
difference
between,
say
conflicting
with
osi
in
conflicting
with
you
know,
some
other
acronym,
that's
in
that's
in
one
corner
of
the
internet.
You
know
osi
is
like
even
though
it's
it's
old,
it's
relevant.
A
I
mean
so
here's
the
other
thing
so
it
you
know
there
was
something
called
cloud
storage,
interface
and
then
there's
another
thing.
You
know
csi
came
up
at
the
same
time
and
it
was
called
cloud
storage
initiative,
csi
and
then
you
know
we
came
up
with
container
storage
initiative
or
interface.
A
A
A
Thoughts,
I
could
be
totally
wrong
about
that.
Just
just
going
off
of
what
I've
seen
in
the
industry,
what
I'm
noticing
and
what
seems
to
be
happening
all
around
us.
F
Yeah,
I
don't
know
it,
you
know
what
I'm
open
to
hearing
more
opinions,
but
but
my
you
know
my
view
is,
I
would
rather
not
have
osi
be
the
acronym
just
because
it
it's
conflicting
with
something.
That's
so
well
known,
and
you
know-
and
I
think
that
like
conflicting
with
any
acronym
that
you
know
that
isn't
my
goal
either
it's
just
not
to
conflict
with,
like
I
wouldn't
choose
tcp
either
tcp
is
you
know
the
same
age,
almost
the
same
exact
age
as
so
as
I,
of
course,
tcp
is
actually
in
use.
F
A
A
F
Actually
ipvc
or
ipv6
is,
is
actually
the
addressing
scheme
from
osi
by
the
way.
But
you
know
at
the
time
that
they
were
choosing
tcp.
They
decided
that
the
addressing
scheme
in
osi
was
too
complex
and
now
we've
adopted
it
with
ipv6,
but.
A
No
but
yeah
I
mean
not
called
osi
naming
ip.
You
know
a
format
right,
it's
again
we're
strictly
talking
about
name.
I
think
it
can
actually
do
us
a
favor
because
of
how
well-known
and
appreciated
osi
is.
If
we
start
going
by
osi,
people
would
associate
us
with
the
same
amount
of
at
least
initially
before
they
find
out
for
real.
What
it's
like.
G
A
Yeah,
if,
if
you
know
I
I
want
to,
I
want
to
go
with
object,
storage
interface,
but
but
again
it's
it's
a
decision
that
we
all
make
together.
A
You
know
during
any
discussion
either
right
now
or
maybe
until
the
end
of
the
next
week,
I'll
bring
it
up
again.
So
people
have
a
chance
to.
I
don't
think
everyone's
on
the
call
today
that
that
generally
join
so
I'll
bring
it
up
again.
The
next
two
meetings
and
we'll
see,
if
others
also
like
this
name.
A
I
think
I
think,
for
the
sake
of
cozy,
for
the
for
the
best
outcome
of
the
effort
that
we're
taking
up,
which
we've
been
doing
for
a
long
time
now,
18
months,
something
like
osi
would
really
help
would
really
make
us
put
us
out.
There
help
us
be
seen
and
noticed
by
others.
A
People
are
going
to
wonder
what
is
this
osi
coming
from
this
kubernetes
group?
It's
definitely
going
to
turn
heads
and,
and
if
you
do
a
good
job
of
implementing
and
so
far
we've
been
doing
a
good
job
and
and
the
product
is
decently
robust.
A
Then
it's
going
to
pick
up
so
I
think
I
think
we
should
keep
this
as
probably
our
front
runner
for
now,
and
we
should
continue
the
discussion
again
because
I
don't
think
such
a
decision
can
be
made
in
one
meeting
in
one
hour
and
and
not
everyone
will
get
a
chance
to
pitch
in.
If
you
just
do
it
in
one
meeting,
they
might
not
be
here
even.
A
A
A
A
I
I
don't
think
this,
you
know,
I
don't
think
there'll
be
too
much
objection
to
this.
I
would
like
to
hear
if
there
are
any.
A
D
A
Okay-
okay,
all
right,
so
that
will
help
a
lot.
Okay.
So
that's
all
I
had
on
the
agenda
for
today
everyone
here.
We
encourage
you
to
start
contributing
to
the
project.
We've
we've
freshly
created
the
project
board,
and
hopefully
that
will
help
people
get
started
with
this
pretty
quickly
and
also
if
you
can't
write
code
right
away.
We
also
appreciate
reviews
on
pull
requests.
A
So
in
order
to
do
that,
just
go
to
all
of
the
individual
projects
and
just
start
watching
them
and
every
pull
request.
You'll
be
notified,
and-
and
you
know
you
can
start
you
can
start
reviewing.
I
can
tell
you
right
now:
the
goal
is
to
get
so.
We've
been
writing
code
outside
of
the
kubernetes
6r,
because
the
repository
repositories
weren't
ready
until
the
kept
was
merged.
So
we
have
a
code
that
is
being
written
and
that
is
ahead
of
what's
there
in
the
repositories
right
now.
A
So
in
the
short
term,
we
want
to
move
all
that
code
to
the
official
repos
and
in
that
process
and
more
people
reviewing
is
good
because
we
want
to
get
the
reviews
done
quicker.
A
A
If,
if
someone
is
started
reviewing
it
try
to
review
the
next
pr
and
and
any
any
chance,
you
know
if
you
have
a
opportunity
to
review
quickly
or
respond
to
comments
quickly
from
who
is
developing
code,
that
is,
that
is
better
than
you
know,
waiting
a
few
days
before
responding
simply
because
until
we
bring
the
current
repo
up
to
par
with
what
code
is
outside,
we
just
want
to
move
things
fast,
because
once
that
is
there,
we'll
have
we'll
have
something
that
everyone
can
test
out
and
run,
and
you
know
play
with.
A
And
if
you
notice
any
issues
or
bugs
please
raise
issues
on
github
itself,
others
can
take
it
up,
we
can
take
it
up
and
it
will
also
serve
as
a
record
of
things
that
need
to
be
done.
B
I
think
you
said
it
right.
I
mean
velocity
is
important
right
now
to
move
the
code
into
the
official
report,
so
we're
trying
to
do
as
much
reviews
as
possible.
Others
can
check
in.
That
would
be
great,
but
this
is
again
the
code
is
lined
up,
so
we
want
to
quickly
merge
some
of
the
important
pieces
of
code
so
that
we
can
move
functionality
and
test
it,
and
I
trade
it
work
your
time
so
yeah.
A
Once
it
is
in
a
testable
way,
you
know
while
integrated
that's
when
that's
when
you
know
people
can
really
start
trying
it
out,
you
know
having
more
meaningful
bugs
raised
or
issues
requested
will
happen
when
people
start
trying
it
out,
and
we
and
also,
if
we
get
to
that
point
where
you
can,
you
know
you
can
run
the
whole
system
as
one
that
will
give
people
something
to
play
with
and
higher
likelihood
that
they'll
come
and
contribute
or
join
us
in
the
meetings
and
such.
A
C
A
All
right,
so
so
that's
it
from
me.
We
have
about
15
minutes
left.
If
there's
anything
anyone
wants
to
talk
about,
we
can
do
it.
If
not,
we
can
continue
on.