►
Description
Kubernetes Storage Special-Interest-Group (SIG) Object Bucket API Standup Meeting - 26 April 2021
Meeting Notes/Agenda: -
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
A
So
the
first
topic
for
today
is
so
I
don't
know
how
many
of
you
know
this,
but
I
work
at
minio
and
last
week
we
so
so
so
mineo
has
been
an
apache
v2
project
for
the
last
five
years
or
so.
But
one
of
the
things
that's
been
happening
is,
as
you
know,
apache
v2
is,
is
a
license
where
anyone
can
make
derivatives
of
your
work
and
then
make
that
proprietary
make
make
their
derivative
proprietary.
A
One
of
the
problems
with
such
a
license
is
one
of
the
good
things
about
such
a
license.
Actually
is
is
that
it
it
drives
adoption
up
and
it's
easy
for
people
to
get
started
with,
but
the
the
flip
side
to
it
is
this
license
almost
goes
against
the
true
philosophy
of
free
software,
where
whatever
software
is
built
or
derivatives
that
are
built
from
this
remain
free
are
supposed
to
remain
free.
A
So
so
we've
been
we've
been
contemplating
this
for
a
long
time
and
and
slowly
we've
been
transitioning
over
our
projects
from
apache
v2
to
agpl,
which
is
a
form
of
a
gpl
license.
Where
you
cannot,
you
know
in
gpl,
you
cannot
create
derivatives
of
a
software,
even
even
by
linking
it
like
code
level
linking,
but
in
agpl
it's
you
can't
even
use
it
over
the
network
and
and
have
it
become
a
commercial
product
or
proprietary
product
without
without
getting
the
author's
license
first.
A
So
anyway,
so
we've
we've
started
transitioning
to
agpl
the
main
io
project.
So,
given
that
and
given
that
we
are
we,
we
have
been
thinking
of
using
min
io
as
our
default
driver.
I
think
we
need
to
have
a
discussion
about
how
to
move
forward
with
that
either.
A
A
We
would
go
this
route,
but
but
to
preserve
the
freedom
of
true
free
software
and
and
to
ensure
that,
like
any
work
that
we
do
is
not
is
not
basically
made
proprietary
and
then
and
then
taken
away
from
the
entire
world.
We've
you
know
we
we've
had
to
go
this
way,
so
if
we
were
to
continue
using
min
io,
the
client
library
and
the
and
the
driver
itself,
the
driver
links
to
the
minio
client
library,
that
is,
that
is
still
apache
v2.
A
However,
if
we
were
thinking
about
using
min
io
inside
our
ci
inside
the
cncf
for
kubernetes
enter
and
testing
ci,
that
is
probably
not
something
that
cncf
might
allow
or
be,
or
maybe
they
might.
We
have
to
figure
it
out
from
what
I
checked,
it
seemed
like
cncf
only
wants
to
work
with
other
like
if
it
were
to
integrate
or
run
some
other
software
in
its
systems.
It
only
wants
to
do
that
with
the
other
apache
v2
projects,
so
so
this
is
something.
B
C
B
D
E
E
A
Yeah,
that's
one
thing
to
discuss
about
the
other.
Is
I've
spoken
to
the
main
io
ceo
and
you
know,
like
he's
happy
to
like
extend
a
free
use
license
for
cncf
purposes,
but
again
we
need
to.
We
need
to
figure
out
if
there's
a
simpler
method
of
this
is
the
right
way.
But
this
is
something
yeah
we
need
to
discuss
so.
B
A
B
Mention
is
if,
if,
for
whatever
reason,
this
turns
into
a
real
problem,
openstack
swift
is
apache
license.
As
far
as
I
know,
and
it's
an
object
store,
it
shouldn't
be.
D
If,
if
we
start
listing
alternatives,
then
well
us
at
scality,
we
also
have
an
apache
licensed
full
s3
implementation.
D
D
So
if
you
have
a
piece
of
software
which
talks
to
mineo
apis
where
using
amin,
io
agpl
licensed
instance,
then
it
kind
of
transfers
to
whatever
is
talking
to
those
api
apis
as
well
in
casual
being
the
mini
of
cozy
provisioner.
A
So
the
client
library,
the
driver
itself,
doesn't
run
in
io.
The
driver
is
just
the
driver.
So
are
you
saying
that
you.
B
A
That
that
kind
of
sounds
right
to
be
honest,
so
I
need
to
verify
that.
Let
me
let
me
talk
to
the
legal
team.
I
I
do
know
what
nicholas
is
saying
that
is
you.
You
cannot
create
proprietary
derivatives
where
it
depends
on
min.
I
o
well
yeah,
so
it
depends
on
mine.
I
o
even
over
the
network
like
it,
can
depend
on.
I
o
as
a
as
a
library
or
over
over
the
network
and
both
are
not
not
allowed,
but
in
this
case.
A
Thing-
and
I
think
you
cannot,
let's
open
it
up,
let's
just
let's
just
open
it
up,
quick
yeah.
B
B
Even
if
you
never
distributed
those
modifications.
The
fact
that
you're
running
it
connected
to
something
would
mean
that
you
would
have
to
distribute
your
modifications
for
minion,
but
you
wouldn't
it
wouldn't
affect
the
license
of
any
of
the
things
that
were
talking
to
it.
It
would
just
say
that
you
have
to
continue
to
open
source
the.
B
A
B
A
Okay,
so
what
it
says
is:
if
you
modify
the
program,
your
modified
version
must
prominently,
offer
all
users,
interacting
with
it
remotely
through
a
computer
network,
an
opportunity
to
see
the
corresponding
source
of
your
version
by
providing
access
to
corresponding
source
from
a
network
at
no
charge
through
some
standard
or
custom
means
of
facilitating
copying
of
software.
This
corresponding
source
should
include
the
corresponding
source
for
any
work
covered
by
version
three
of
gnp.
B
It
means
that
minio
and
its
derivatives
have
to
be
agpl,
but
the
the
you
can
still
write
things
that
talk
to
it
with
whatever
license
you
want.
A
A
B
Think
it's
sorry.
C
Go
ahead,
go
ahead
so
from
what
I
know
about
this,
and
I
posted
this
tldr
legal
one.
If
you
want
to
take
a
look,
it's
it
just
mainly
says
what
you
have
to
do
if
you
use
this
license,
so
the
license
just
have
terms
of
what
you
have
to
do,
and
and
as
soon
as
you,
you
know
you
meet
these
terms.
It's
not
that
you
have
to
implement
a
gpl
in
every
dependent
work
and
adopt
it.
C
It's
just
that
if
you
make
changes
to
the
software
specifically
to
its
code
and
and
you
know
that
it
doesn't
extend
over
the
network
so
that
you
become
a
gpl
just
by
consuming
it
over
the
network,
but
it
does
mean
that
if
you
are
providing
a
network
service
of
that
sort,
you
have
to
distribute
your
source.
That's
that's!.
C
C
Hey
you
can
just
use
unmodified
version
as
well.
It's
so
it's
it's
not
as
restrictive.
It's
mainly
just
saying
that
you
have
to
use
an
open
source
so
that
the
source
and
any
kind
of
small
modifications
you
make
to
make
it
better
still
becomes
open
source.
That's
that's
the
main
thing,
but
other
than
this.
I
you
know
it's
pretty
permissive
right,
yeah,
okay,
so
there.
A
Okay,
so
that's
the
difference:
yeah
mongodb
yeah
anyway.
So
that's
that's
good
to
hear
so.
Is
there
anyone
we
should
follow
up
with.
A
Maybe
talk
to
talk
to
legal
somewhere,
I
mean
I
can
do
this
one
thing
I
can
talk
to
the
legal
at
minao.
Is
there
someone
at
cncf
or
someone
we
should
talk
to
sheng?
Would
you
know
or
anyone
else.
E
A
Okay
sounds
good
all
right.
Let's
do
that
just
to
make
sure
I
think
I
think
overall,
it
looks
fine
based
on
what
we
just
discussed,
but
it
would
be
good
to
just
get
the
official
okay
on
this
all
right.
So
this
is
let's
move
on
to
the
next
thing.
Let's
see.
A
So
did
we
did
we
reached
a
conclusion
on?
Did
we
reach
a
conclusion
on
credential
rotation?
Oh
yeah,
we
said
we
said
we'll
we'll
we'll
deal
with
it
after
alpha
right,
yeah,.
B
A
Right
now,
that's
good
all
right,
so
I
want
to
focus
a
little
bit
on
coding
today
in
a
sense,
in
a
little
bit
on
development,
just
get
the
plan
out
there
in
terms
of
where
we
are
and
what
we've
got.
I
believe.
C
Can
I
ask
a
quick
question
on
the
previous
topic
just
to
raise
the
point,
but
we
don't
have
to
spend
time
on
it?
Do
you
think
that
this
driver's
repository
do
you
think
it
will?
It
will
want
to
accept
more
drivers
so
that
it,
the
ci
you
know
covers
more?
Is
that
kind
of
something
that.
C
E
A
A
E
E
E
Yeah
yeah,
if
it's
a
ripple
under
you,
know
yeah,
then
we
have
that
yeah.
So,
like
the
this
other
repos
under
kubernetes
csi,
we
can
we
set
up
cis
for
those
as
well.
A
E
A
E
It's
I
think
it's
I
was
actually
I
don't
know
what
was
the
reason.
Maybe
some
of
the
general
test
testing
framework
does
not
fit
or
something
they
are
actually
doing.
Something
like
they
use
the
github
actions,
so
yeah.
A
Okay,
yeah
they're
using
github
actions,
yeah
got
it
okay,
very
nice,
so
so
is
that
how
we
are
planning
on
expanding?
Are
we
thinking
of
having
a
driver
per
vendor
or
having
a
driver?
Well.
E
It's
not,
I
don't
think
we
are
going
to
open
up
for
all
the
vendors,
because
if
you
look
at
it,
if
you
look
at
the
css
driver,
you
know
how
many
vendors
have
cso
drivers,
we're
not
going
to
have
them
all
hosted
here,
so
these
are
all
for
with
the
sample
or
certain
criterias
that
we
can
host
them
here.
Normally
we
don't.
We
don't
host
the
driver
under.
D
E
So
yeah,
so
that's
actually
one
thing
that
we
don't
really
have
like
a
a
requirement
for
like
certification
type
of
thing.
We
don't
really
have
that
because
we
don't
really
have
bandwidth
to
really
do
that,
but
but
we
do
have
tests
available,
so
drivers
can
run
those
tests
or
css
tests.
Those
are
for
csi
conformance
cs
back
conformance,
and
then
there
are
also
e2e
tests.
Driver
can
actually
download
those
etv
tests
and
run
their
own
test.
Their
own
driver.
E
Maintenance
from
the
I
actually
ben.
You
also
know
that
I
I
actually
I
think
it's.
Yes,
it's
a
very
high
maintenance.
If
you
look
at
the
like
openstack
the
cinder
test
and
then
it's
just
it's
kind
of,
I
think
it's
it's
too
difficult
to
maintain
that.
D
To
test
etc
as
part
of
the
kubernetes
desktop
harness,
so
a
cozy
provisioner
is
the
responsibility
likely
of
the
vendor
of
the
backing
object,
storage
system,
one
could
say:
there's
a
couple
of
open
source
storage
systems
for
which
a
driver
could
be
developed
as
part
of
kubernetes
6,
but
in
general
it's
not
now
having
a
test
harness
or
a
test
suit.
You
can
use
as
a
vendor
to
validate
your
driver
is
likely
doing
the
right.
Your
provisioner
is
likely
doing.
The
right
thing
is
super
valuable.
Even
if
it's.
A
A
What
she's
saying
is,
while
we
completely
agree
on
that,
it's
about
what
it
takes
to
keep
that
up
to
date
and
also
what
it
takes
to
you
know
work
with
everyone
on
this,
so
the
amount
of
effort
it
takes
to
do
this
is,
is
very
large,
very
high,
and-
and
I
think
we
should
do
it,
but
it's
just
we
should
we
should
look
at
when
is
the
right
time
to
do
it,
maybe
because
again
we're
not
so
complicated
that
we
need
a.
We
need
an
automated
test
suite
to
understand.
A
If
you're
doing
the
right
things
I
think
to
begin
with,
we
can
we
can
have.
We
can
just
do
manual
testing.
At
this
point.
We
have
a
very
small
project
at
this
point,
and-
and
things
like
this
will
come
up
as
we
get
bigger
and
bigger,
and
at
that
point
we
can
actually
iteratively
step
by
step
at
it
and
it'll
be
much
more
manageable.
If
we
did
it
that
way
than
to
start
it
as
a
as
a
large
project
and
then
try
to
drive
it
like
at
one
point
in
the
future,.
C
B
A
C
E
Yeah,
so
I
wonder
maybe
like
if
we
have
some
tests
and
have
driver
test
that
and
just
submit
some
results
be
better.
I
do
not
think
we
if
we
have
like
continuous,
like
cse
pipeline,
like
constantly
testing
all
the
vendor
drivers.
I
don't
think
that
is
manageable
at
all.
It's
just
doesn't
make
sense
to
me.
A
Right
right
right,
I
mean
we'll
we'll
lose
the
sorry
go
ahead.
A
Oh,
I
was
saying
if
we
were
to
go
that
route
of
doing
it,
for
every
driver
like
that,
we
lose
the
whole
benefits
of
micro
services
that
we
that
we
all
are
going
towards,
because
because
everyone
will
have
to
converge
on
one
instance
of
the
api
and-
and
you
know,
we
can't
move
the
api
forward.
If
all
the
vendors
don't
conform,
because
all
of
the
tests
will
fail.
C
You
know
you
know
what,
but
something
that
could
be
valuable
in
that
scope
could
be
if,
if
we
can
all
collaborate
on
creating
the
like
the
test
software,
like
a
test
repo
for
a
driver,
that
that
would
be
like
something
that
all
vendors
would
be
able
to.
You
know
update.
I
guess.
C
Right
so,
if,
if,
if
the
entire
test
that
runs
in
the
ci
for
validating
cosy
right
with
with
min
io
and
the
ci,
if,
if
that
can
be
reused
for
testing
any
driver,
and
that
would
be
a
valuable
upstream,
you
know
reaper
to
contribute
and
having
yeah.
E
A
I
see
okay
yeah.
That
makes
sense
so
so
guy
what
you
said.
I
think
I
think,
even
the
tests
we
write
from
an
io
only
the
provisioning
steps
would
be
different.
I
think
in
terms
of
the
actual
testing
itself,
it
would
only
talk
the
s3
api
in
terms
of
the
bucket
consumption
side
and
in
terms
of
yeah.
A
In
terms
of
provisioning
side,
I
mean
again,
it's
s3
api
for
bucket
creation
and
just
bucket
specific
calls
are
s3
apis
only
difference
between
when
I
own
s3
would
be
in
in
provisioning
access
credentials,
because
in
in
s3
you'd
be
calling
iam
service
in
s3
and
when
I
you'd
be
calling
min,
I
o
admin
service.
So
that's
that's
one
point
of
difference,
but
yes,
I
completely
agree
that
any
test
we
write
should
be
for
should
be
as
generic
as
possible
and
and
if
possible,
we
should.
A
Like
you
said,
if
everyone
contributes
to
it,
we
we
would
definitely
have
a.
We
would
definitely
have
a
good
base
to
work
off,
of
and
and
also
if,
if
mineio
used
the
generic
portions
of
this
of
this
test
harness,
then
it
would
show
other
vendors
new
vendors.
How
to
how
do
you
take
the?
A
How
do
you
take
the
generic
portion
and
then
modify
it
to
work
with
their
their
system?
Yes
sounds
good
yeah.
We
should
talk
about
that.
Actually
now
now
now
that
we
have
so,
first
of
all,
even
before
we
go
there
chris
do
you
have
a
demo
to
show
us
if
you're
there.
A
So
so
chris
is
so
krish
has,
you
know,
done
the
work
on
he's
written
a
tiny
application,
just
like
a
demo
application
that
relies
on
a
bucket
being
proved
provisioned
via
cosy.
A
So
it's
it's
a
working
demo
which
goes
over.
You
know
in
the
demo.
He
shows
the
bucket
life
cycle
operations
like
creation
of
bucket
access,
bucket
access,
and
all
that
and-
and
he
shows
some
I
believe,
some
object
being
uploaded
to
a
bucket
and
and
just
the
whole
life
cycle.
A
The
whole
idea
behind
this
demo
was
when
we,
when
we
originally
started,
one
of
our
main
goals
was
to
show
a
demo
the
idea
behind
it
was
by
by
putting
that
as
the
goal
we
would
be
forced
to
build
the
entire
end-to-end
architecture
rather
than
just
focus
on
one
at
a
time
and-
and
it
is,
it
is
definitely
paid
off
in
the
sense
that
we
have.
We
have.
We
have
the
different
components:
working
to
the
extent
where
we
can.
A
We
can
rely
on
it
to
provision
buckets
revoke
access,
create
access
and
delete
buckets
again
again
with
minah
for
now
so
yeah
chris
will
show
to
us
in
a
little
bit
right
after.
A
If
we
have
time
we
should,
we
should
talk
about
this
testing
hardness
now
that
I
now
that
you've
brought
it
up,
not
really
a
testing
harness
but
kind
of
a
something
generic
as
a
starting
point,
I
think
it'll
it'll
it'll
address
what
nicholas
asks
for
as
well,
and
also
what
guy
was
saying,
because
by
doing
this,
we
will
end
up
in
a
position
where
anyone
who
writes
a
driver
can
kind
of
get
started
with
and
and
see.
You
know
if
the
driver
is
doing
well.
A
If
it
satisfies
the
api
or
not
so
so,
seeing
you
said
you
know,
the
cinder
tests
were
very,
very
complicated.
To
maintain.
Was
that
because
of
like
you
know,
in
openstack
cinder,
there's
a
lot
of
wiring
involved
in
how
like
a
cr,
you
know
a
volume
is
like
wired
into
a
into
an
operating
system
or
into
into
a
host
or
vm.
D
The
main
issue
with
cinder
was
I,
from
my
perspective,
at
least,
that
the
vendor-specific
drivers
were
part
of
cinder
and
shipped
with
cinder
and
as
a
vendor
you
had
to
when
you
wanted
to
make
a
code
change.
It
had
to
be
integrated
in
cinder
and
the
sender
ci
was
set
up
such
that
every
vendor
who
wanted
to
have
a
driver.
D
Part
of
cinder,
had
to
have
its
own
ci
system
in
place,
which
would
then
report
for
every
single
code
change
not
only
to
your
own
drive
but
to
any
driver
or
any
of
the
common
code.
We
need
to
your
ci
will
be
triggered
and
we
need
to
report
back
success
or
failure,
and,
given
that
there
were,
I
don't
know,
50
to
100
drivers
in
cinder
that
caused
a
lot
of
flakiness,
the
third-party
ci
being
unavailable,
rtcis
being
unreliable
themselves
and
always
impacting
any
code
change
throughout
the
cinder
source
code.
D
B
Cinder
was
that
with
the
because
the
drivers
themselves
were
part
of
the
project
in
principle,
anyone
should
be
able
to
modify
them,
not
just
the
vendor
that
owned
them.
But
if,
if
some
third
party
wanted
to
go
fix
a
bug
and
say
the
netapp
driver
like
they
would
need,
they
might
not
have
netapp
hardware
to
go
to
go
test
their
their
fix
on
and
so
netapp
was
required
to
run
ci
so
that
we
could
tell
them
if
their
change
broke
us
or
not,
and
then
in
principle.
B
D
Someone
made
a
change
to
the
netapp
driver,
then
the
emc
tests
or
ci
had
to
run
and
run
and
succeed
as
well
yeah,
but.
B
Yeah
yeah
yeah,
I
mean
that
was
that
was
a
whole
nightmare,
but
I
mean
the
the
principle
of
it
kind
of
made
sense
if
you
bought
that
it
was
a
good
idea
to
have
the
drivers
be
in
the
tree
in
the
csi
world
in
the
cozy
world,
every
driver
is
its
own
project
separate.
So
there's
all
of
these
requirements
don't
make
any
sense,
but.
A
Also,
you
know
doing
it
that
way
goes
against
the
whole
philosophy
of
micro
services
in
the
microservices
is
not
just
about
breaking.
You
know
software
into
smaller
chunks,
it's
about
breaking
it
along
along
boundaries
of
you
know,
operations
like
people
and
and
like
teams
that
are
developing
it.
You
basically
make
them
independent
to
one
another
and,
and
the
only
point
of
integration
is
the
api.
B
B
E
B
E
B
A
Oh
so
so
one
is,
I
think
there
should
be
some
mechanism
to
make
sure.
A
Let's
say
we
make
a
change
in
and
cozy
api.
Let's
say
we
we
add
a
new
or
we
go
and
remove
a
field
or
or
we
change
how
a
bucket
is
provisioned
like
we
stop
looking
for
bucket
name
is
empty
or
not,
and
you
know
so
so
any
driver,
that's
that's
importing
this
code.
Any
driver-
that's
based
on
this
library,
should
should
know
that
you
know
they
can't
send
in
empty
bucket
names
anymore.
A
So
so
they
should
fail.
But
it's.
B
B
You
update
the
spec
itself
to
say
now
we're
going
to
add
this
new
optional
parameter
or
whatever,
and
then
you
change
the
contract
by
changing
the
spec,
but
as
but
anyone
who
compiles
with
the
older
version
of
the
spec
will
get
the
older
behavior
and
as
long
as
every
change
is
backwards
compatible.
You
never
ever
have
a
problem,
and
this
is
how
csi
works
right.
A
D
Trouble
compilation:
there
is
a
bunch
of
things
in
the
csi
spec
related
to
which,
which
error
you
have
to
report
as
a
driver.
In
this
particular
scenario,
this
particular
rpc
call
must
be
idempotent,
etc,
etc,
which
are
not
which
are
part
of
the
the
spec
text,
but
not
part
of
the
csi
bindings.
The
I
mean
the
the
compiled
protofile
into
some
vocode
or
whatever.
So
this
is
where
csi
test
comes
in
to
actually
validate
that
the
driver
behaves
according
to
the
specification,
not
only
in
which
rpc.
A
B
Authors
to
basically
sanity
check
their
their
driver,
yes,
and
and
and
having
worked
with
the
csi
sanity
tests
they're,
they
leave
something
to
be
desired:
they're
they're,
they're
too
loose
in
some
places
and
too
strict
in
other
places.
B
I
mean
they're
better
than
nothing
by
a
long
shot,
but
they
could
they
could
they
they
could
be
way
better.
I
wouldn't
recommend
someone
writing
a
csi
driver
use
those
religiously.
I
might
run
it
once
and
look
for
any
problems
and
say:
okay,.
A
You're
saying
if
it
found
something
you
know
it
could
be
useful,
but
if
it
didn't
find
anything,
it
doesn't
mean
that
you
know
you're.
Okay,.
B
Right
and
sometimes
it
finds
stuff,
that's
that's
bogus
just
because
of.
A
A
And
yeah
I
mean
it
goes
to.
You
know,
support
what
shin
was
saying.
I
think
yeah.
This
is
kind
of
a
convenience
for
sanity
testing
kind
of
thing.
If
it's.
B
C
C
That's
mainly
my
my
my
point,
I
mean
I
think,
if
we
invest
in
it,
I
think
cozy
will
be
more
successful,
but
you
know
if,
if
we,
you
know
kind
of
end
up
doing
each
other
like
each
own
ci
for
for
the
driver,
we'll
get
something
more
variant
across
the
certification
that
cozy
provides
so
yeah.
A
C
Right
for
the
driver,
you,
if,
if
you
are
trying
to
communicate
to
the
driver-
and
you
want
to
test
both
cosy
and
the
driver,
the
specific
driver
through
that
and
in
an
end-to-end
test,
you
would
impersonate
an
admin.
Administrator
kind
of
you
know:
operations
and
user
operations
right
and
like
check
it
from
both
views
of
what
cozy
provides,
because
we
do
specify
something
of
the
behavior
in
the
documentation.
But.
B
B
C
You
know
look
for
that.
I
would
just
say
that
it
makes
it
well
when
you
get
less
variance
about
how
this
behaves,
and
this
becomes
more
of
a
you
know-
good
routing
layer
for
for
vendors
and
to
to
use.
Then
I
think
you
know
vendors.
A
Point
faster
confidence
for
sure
yeah.
It
improves
confidence
that
that
they've
written
it
right
and
you
know
they
they'll-
also
know
that
whatever
the
shipping
is
is
con,
you
know
is
in
line
with
what
others
are
also
doing.
So
I
I
agree,
but
guys.
This
is
something
we
should
look
at
when
we
have
v1
right.
That's
how
you're
thinking
about.
C
Yeah
yeah,
I
I
think
I
yeah,
I
didn't
mean
it
to
be
something.
That's
so
super
important
at
this
point
right
right.
F
F
A
Okay,
so
let
me
yeah,
let
me
do
that.
Okay,
give
me
a
second,
so
so
we're
doing
a
talk.
Okay,
let
let
me
first
ask
for
permission
so
shane.
This
is
a
this
is
for
the
talk
we're
submitting
to
cndm
days.
Oh.
E
A
A
Part,
oh:
does
this
show
in
full
screen.
F
A
Oh,
like
that,
okay,
if
you
didn't
find
it,
I
think.
A
Yeah
yeah,
I
know
what
you
mean,
okay,
so
it's
here
yeah
there.
It
is.
F
A
Yeah,
I
think
it.
A
Yeah
should
be
the
name
of
demo
sequence,
so
we
tried
sharing.
So
we
both
created,
accounts
on
on
creative
cloud
and
okay
christian
true,
so
it
has
a
git
like
interface,
for
I
don't
see
it.
A
F
F
E
A
No,
we
we
used
adobe
premiere
pro,
oh
okay,
yeah.
It
should
be
it
should.
Maybe
it
was
the
internet
connection.
I
don't
know
because.
F
Yeah
yeah,
I
think
it
might
be
the
are
you
sharing
your
screen.
If
you
share
the
application,
I
think
you
can
share
the
audio
and
the
application
right
like
whatever
in
zoom.
When
you
share
your
screen,
you
can
select
to
share
an
application.
I
think
when
you
share
an
application,
it
does
the
audio.
A
I
don't
see
it
oh
anyway,
so
so
so
so
shane
and
and
ben
yeah.
The
audio
sounded
fine
on
my
side.
So
it's
probably
zoom.
That's
having
the
issue
well,.
E
F
Okay,
so
at
this
point
we're
just
going
through
the
manifests.
So
this
is
the
bucket
request
and
bucket
access
request.
F
Nothing
like
we're
not
doing
any
like
fancy
configuration
here
just
to
keep
things
simple
and
then,
within
the
actual,
like
workload
pod,
we
define
our
volume
for
our
secret,
so
this
will
be
where
the
actual
credentials
are
mounted,
and
so
once
we
create
our
classes,
I
think
yeah
then
we're
going
to
create
our
bucket
access
for
question
bucket
request
and
pretty
quickly
both
like
the
corresponding
bucket
and
bucket
access
are
created,
and
so
we
can
check
the
status
of
the
bucket,
and
we
can
see
here
that
the
bucket's
available-
and
you
know
the
bucket's-
been
provisioned
in
the
backing
provider.
F
Some
displacement,
I
o
right
and
we
can
do
the
same
thing
for
the
bucket
access.
So
you
can
see
here.
Access
has
been
granted
and
and
there's
a
minted
secret.
That's
been
created
and
there's
like
a
secret
reference
there
for
the
the
node.
E
F
Yeah,
and
so
now
we
create
our
actual
like
workload
pod,
and
so
we
can
check
min
io
locally.
I
just
forwarded
the
port,
and
so
we
can
see
that
our
sample
bucket
has
been
created.
That
was
the
that
was
the
id
we
chose,
and
so
now
this
is
the
actual
web
application.
That's
connected
to
minio
through
a
back
end
running
in
the
running
in
the
cluster,
so
you
can
just
select
images
to
upload
here.
F
It's
like
a
instagram
clone.
Basically
right
so
just
select
the
image
you
know
crop
it
into
a
square
and
upload
it,
and
so
we
can.
We
can
set
our
name
for
the
picture
as
well.
So
in
this
case,
let's
do
it
gophers
or
something
oh,
go.
First
plus
kubernetes,
my
bad,
and
so
if
we
go
back
to
the
homepage,
you
can
see
that
we
just
like
you,
can
see
all
the
pictures
here,
and
so
I
think
the
rest
of
the
demo
is
just
me.
F
You
know
adding
some
more,
adding
some
more
fun
pictures
and
like
everything's
handled
with
just
min
io
and
and
the
bucket
that
we
and
the
bucket
access
that
we
provisioned
with
with.
F
A
Any
thoughts
on
this,
so
if,
if
anyone
wants
to
set
up
the
whole
system
or
if
anyone
wants
to
do
the
same
steps
as
this
demo,
you
can
obviously
reach
out
to
us.
But
but
at
this
point
I
think
I
think
I
think
more
or
less
people
should
be
able
to
do
it
already
right.
F
Yeah
should
be
able
to,
I
think,
with
the
version
of
the
csi
adapter.
That's
on
master.
I
believe
that
should
work.
A
Yeah
yeah,
that's
good,
and
but
there's
one
change
that
was
missing
right.
The
revoking
access
or.
F
Yeah,
so
the
version
of
grant
access
that
we
used
here
is
kind
of
like
naive
implementation,
so
I
think
janice
is
still
working
on
the
actual
version
of
grant
access,
which
will
like
create
a
new
account
and
stuff,
and
so
that
should
that
should
be
good
in
in
in
a
few
days,
I
think,
whenever
that's
emerging.
A
Yeah
so
yeah,
that's
that's
where
we
are.
We
just
have
to
merge.
There's
already
a
pr.
We
just
have
to
go
through
the
review
cycle
and
merge
it
for
the
for
the
entire
end
to
end.
You
know
first
implementation
of
the
entire
intern
to
work.
It's
definitely
demo
ready.
I
wouldn't
yet
call
it
production
ready
or
anything,
but
but
it's
getting
there.
A
I
would
encourage
everyone
to
try
it
out.
Try,
try
it
either
with
whatever
driver,
you're,
writing
or
try
it
with
with
mine.
I
o
see
how
it
goes
and
report
back
any
errors
that
you
see
and-
and
if
you
have
any
questions
or
anything
you
know
feel
free
to
ask
right
now
or
later
is
fine
too.
C
A
Okay,
all
right,
that's
it
for
today,
on
on
thursday
again,
I
just
want
to
go
over
some
of
the
development
tasks
that
are
remaining
again
would
would
appreciate
any
help
from
anyone
who's
who
has
some
cycles
to
contribute?
Hey.
E
Do
you
do
you
know
the
1.22
schedule
is
out,
yeah
enhancement.
Freeze
is
actually
let
me
check.
How
was
the
time
it's
may
13th.
So
it's
the.
A
Got
it
okay,
so
so
sushi,
so
I've
reached
out
to
tim
so
last
week
he
hasn't
responded
yet
so
I'll
reach
out
to
him
again
today.
Is
there
anything
else
we
can
do
to
speed
it
up.
E
I
think
it's
just
reaching
out
to
him.
I
guess
I
mean
I.
I
don't
have
a
any
better
way
you
can
pin
inside.
He
can
also
help.
A
Okay-
and
I
think
last
time
someone
mentioned
there's
a
there's-
a
cap
reviewer
other
than
the
api
reviewer-
is
that
right.
F
E
A
Okay,
yeah
I'll
I'll
reach
out
to
tim
again,
we'll
see
we'll
see
what
he
says:
okay,
all
right!
Okay,
that
that's
it
from
me!
Is
there
anything
else.
I
want
to
talk
about.
B
I
just
wanted
to
mention
that
my
group
at
netapp
is
having
a
so-called
hackathon
next
week
and
I
got.
B
To
sign
up
to
write
a
cozy
driver
in
our
hackathon
wow
so
next
week,
we'll
be
we'll
be
doing
that.
A
Great,
that's
great,
let
us
know
how
it
goes.
I
mean,
let
us
know
if
you,
if
there's
anything,
we
need
to
do
say,
for
instance,
anything
in
the
library,
that's
missing
that
might
prevent
you
from
you
know
finishing
it
you
you
can,
let
me
know,
but
but
I
think
right.
C
A
We're
in
a
good
shape,
we
have,
we
have
enough
that
you
can
show
a
full
demo,
it
will
be.
B
Netapp
has
two
products
that
that
are
object,
stores
at
least
two.
I
guess
there
could
be
more,
I
don't
know
about,
but
but
we
plan
on
supporting
two
of
them
eventually
in
trident
and
so
we're
gonna
prototype
it.
It's
gonna
be
next
monday,
tuesday,
wednesday
we'll
be
working
on
this
stuff.
So
I
might
miss
the
next
monday
meeting
as
a
as
a
result
of
that,
but
but
maybe
maybe
they'll
have
some
questions
and
I'll
show
up
and
ask
my
questions
in
the
next
monday
meeting.
A
All
right,
good
luck
with
that.
If
you
don't
see
you
well,
we'll
see
you
on
thursday,
anyways
yeah
yeah
I'll
be
around
all
right,
so
all
right,
good
meeting,
let's
meet
again
on
thursday,
I
want
I
want
to
do
a
little
bit
of
development
planning
on
thursday
and
then
and
then
we'll
get
into
the
remaining
things
that
we
have.
I
yeah
we
have
it
written
down
in
the
project,
but
I
couldn't.
A
I
couldn't
quite
tell
you
off
the
top
of
my
head
that
the
things
that
we
need
to
focus
on
there's
a
lot
remaining,
so
yeah
all
right
talk
to
you
all
on
thursday.
Thank
you.
Everyone.