►
From YouTube: Kubernetes SIG Cloud Provider 2019-10-02
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
All
right,
hello,
everyone
today
is
October
2nd.
This
is
the
bi-weekly
kubernetes
2
cloud
provider
meeting
reminder
to
add
your
name
on
the
pending
list
for
today's
meeting
in
the
agenda.
So
we
just
get
started
straight
to
the
agenda.
First
action
item
for
the
cig
is
getting
a
replacement
for
Chris's
speaking
spot
for
our
sick
and
so
I
believe
Chris
and
Fabio,
or
signed
up
to
do
the
sick,
intro
and
with
Chris's
new
transition.
A
B
The
so
the
the
the
deadline
that
they
stated
for
the
speaker
registrations
was
September
30th,
but
I
got
an
email
last
week.
Saying
that
hey,
you
need
to
like
tell
us
if
you're
speaking
by
October
4th
so
I
I
filled
out
the
registration,
because
I
already
have
a
past
for
some
of
the
for
the
track.
Sure
work
I
did
last
year,
but
I
would
like
to
before
the
end
of
the
week,
get
a
note
to
Nancy
as
to
who's
going
to
replace
me
on
those
speaking
sessions.
B
B
B
A
B
And
the
nice
thing
about
these
sessions,
too,
is
that
they're
there's
some
presentation
going
on,
but
there's
also
a
lot
of
community
engagement.
So
it's
it's
not
just
about
the
speaking,
but
it's
also
about
engagement
with
the
rest
of
the
sig,
so
they
wind
up
it
winds
up
being
a
very
gentle
entry
into
cubed,
cube,
Khan
speaking
I
would.
D
D
E
Say
I
think
it
would
be
great
if
someone
like
you
took
it,
Nick
I
mean
we
want
to
show
more
faces
than
just
you
know.
Fabio
myself
and
Andrew
we've
been
doing
a
lot
of
these
and
it
it's
gonna
help
our
sig
to
show
that
we
do
have
a
broader
reach,
so
I'm
I
will
certainly
+1
you're
doing
it
Nick.
That's
not.
D
B
A
See
so
I
will
follow
up
with
Fabio
and
let
him
know
that
you're
interested
because
he's
the
other
Co
speaker
and
and
yeah
I
think
there
is
anyone
else
in
this
league
that
is
interested
to
ping
us
offline
and
will
coordinate,
but
tentatively
I'm.
We
have
Nick
Turner
set
to
speak
there
cool
all
right
next
item
on
the
list.
Walter
did
I
put
this
or
did
you
write,
I'm
losing
track
I
figured.
A
So
so
there
was
a
call
to
action
from
the
community
to
go
ahead
and
go
through
our
owners
files
and
remove
folks
who
have
been
inactive
but
also
add
folks,
who
have
been
active
but
not
been
promoted
to
either
reviewer
or
approval
er.
But
on
that
note
like
it
might
be
a
good
time
to
also
add
an
official
owner,
alias
for
our
sake.
A
E
A
couple
of
things
so
one
I'd
like
to
make
sure
we
have
at
least
two
owners
in
every
owners
file,
also,
if
you
aren't
quite
feeling
like
you're,
ready
to
be
an
owner.
Yet
please
at
least
make
your
ask
yourself
to
be
a
reviewer.
I
reviewers
is
a
great
way
to
get
some
practice
into
being
an
owner.
If
you
don't
feel
quite
ready
and
the
more
owners,
we
have
the
more
community
commitment
we
have
the
healthier.
This
seg
is
going
to
be.
A
E
All
right,
so
this
wasn't
definitely
mine.
That's
yeah
did
my
name
there.
So
I
think
it
would
be
really
valuable.
I
I
think
I've
said
this
before
for
us
to
try
and
generate
an
official
cloud
provider
representatives
list
one
you
know
so
there's
a
couple
of
questions.
One
is
where
would
we
host
it
somewhere?
That's
discoverable,
I.
Imagine
that
quite
a
few
people
are
also
going
to
wonder
what
are
the
responsibilities
of
the
cloud
provider
representative.
E
A
first
good
example
of
I
think
the
value
is
a
little
project
that
Andrews
been
working
on,
which
is
a
tracking
sheet
for
how
the
out
of
tree
or
excuse
me
how
the
infiltrate
providers
are
working
on
getting
out
of
tray.
So
obviously
we
need
people
who
are
working
on
these
projects
to
be
able
to
tell
us.
Is
this
piece
done?
Where
are
you?
Are
you
in
danger
of
now
making
it
by
a
certain
deadline?
A
Yeah,
so
this
is
a
so
yeah
like
just
for
some
context.
As
focusing
you
know,
we
have
the
long-running
effort
trying
to
get
things
out
of
tree
and
it's
excruciatingly
Li
slow,
but
we're
making
progress
slowly.
So
one
thing
we
found
that
could
be
helpful
is
just
kind
of
having
some
sort
of
you
know.
Official
tracking
sheet,
that
kind
of
says
the
current
state
of
things,
which
requires
involvement
from
all
the
cloud
providers
to
actually
report
on
things
so
like
so,
for
example,
Richard
that
guy
I
could
I
didn't
know
how
to
fill
out.
A
If
I
give
you
added
access,
yes,
but
if
not,
please,
let
me
know,
and
I
can
add
you
to
the
tracking
sheet
to
just
update
the
IBM
parts,
but
also
like
other
folks
and
other
providers.
If
any
information
here
is
inaccurate,
please
reach
out
and
we
can
get
this
filled
up,
so
we
can
kind
of
just
keep
a
tally
of
where
things
are
and
what's
behind,
what
needs
more,
what
needs
to
be
prioritized
for
future
releases
and
whatnot.
A
C
E
C
E
Well,
this
is
so.
What
are
the
things
that
I
think
beyond
the
value
that
the
and
you're
just
demonstrated?
Some
other
things
when
we
get
done
would
be
be
getting
the
entry
out
of
tree
effort?
I,
don't
think
the
cig
just
wraps
up
I
think
there
are
going
to
be
a
lot
of
issues
that
we're
going
to
want
to
continue
addressing
some
unpleasant,
but
but
definitely
vital
ones
include
things
like
once
all
the
cloud
providers
are
out
of
tree.
We
need
to
start
having
signal.
C
E
You
know
orb
and
the
key
go
in
either
direction,
but
we
need
points
of
contact
for
hey
who
do
I.
You
know
as
an
example,
since
I'm
Google
I
noticed
that
I
made
some
change
to
the
way
node
names
are
and
suddenly
all
of
G
K
ease,
test
grid
has
turned
red
and
I,
don't
even
know
how
to
run
those
tests
because
effectively
only
someone
at
Google
can
run
them.
Can
you
help
me
the
who
made
this
change
understand?
E
C
Ok,
you
convinced
me
of
the
value
of
having
such
a
thing,
but
my
secondary
reaction
is
maybe
that
it
not
even
constrain
itself
to
just
a
top
level,
but
maybe
offer
an
option
of
people
to
itemize
it
for
specific
areas
of
particular
providers
like
storage
and
CSI.
Maybe,
even
though
this
isn't
really
under
the
realm
of
say
cloud
provider,
even
cluster
API
is
some.
You
know
an
itemised
thing
where
having
a
name,
there
could
potentially
be
valuable
and
granted.
You
could
go
put
it
under
the
other
sig
that
owns
cost
or
API.
C
E
C
Well,
for
some
cloud
providers,
maybe
the
name
is
a
is
the
same
person
or
people
for
every
subject.
I
mean
it's
possible,
but
for
others
that
are
a
little
more
complex.
I
can
easily
see
where
an
itemized
list
would
make
more
sense
and
the
itemized
Leclaire
option
for
a
cloud
provider
to
itemize
if
they
want
yeah.
E
A
Six
kind
of
just
inherit
this,
like
role
of
representing
the
cloud
providers
and
seeing
in
hat
and
providing
some
sort
of
feedback,
loop
feedback,
loop
or
some
sort
of
input
on
what
changes
are
happening
in
core
and
how
that
impacts,
their
cloud
provider,
implementation,
but
I
think
that's
not
always
the
case
like
that.
We're
not
really
getting
that
participation
that
we
want,
and
so
I'm
not
opposed
to
this
idea
as
well.
E
D
E
It
seems
like
it
would
be
good
to
at
least
have
one
person
from
Amazon
on
that
list:
I
mean
and
yeah,
so
and
and
and
then
there
are
others
I
mean
at
least
we
have
a
very
healthy
Amazon,
one
I
think
for
some
of
the
other
cloud
providers.
They
may
have
been
small
enough
that
they
never
really
got
a
functioning.
E
You
know
sick
of
their
own,
and
so
we
will
be
missing
a
lot
of
those
names
and
so
I
think
this
may
be
an
opportunity
for
some
of
the
smaller
cloud
providers
to
let
us
know
who
we
should
contact.
You
know,
we've
all
been
through
the
floor
at
Q,
Khan
I
know
there
are
a
lot
of
small
cloud
providers
out
there.
E
E
G
It
is
kept
deadline
on
the
15th
and
just
like
every
year,
it's
a
slightly
shorter
cycle
due
to
winter
holidays
in
Cube
cone
and
that's
about
all
I
had.
My
next
question
is
what
interesting
Steele
have
coming
up,
but
I
think
that's
exactly
what
the
next
point
is
going
to
be
about.
So
go
quiet
now
and
listen
more.
A
E
I
I
The
the
biggest
blocker
we
still
have
is
around
the
way
that
cloud
controller
manager
interacts
with
API
and
the
Iser
API,
and
so
we're
still
working
on
several
ways
to
solve
our
API
throttling
challenge,
and
you
know
I,
guess
I'm
not
up
to
terribly
optimistic
that
we
will
solve
those
before
enhancing
trees.
At
which
point
we
would
you
know
if
we
did
then
I
think
it
would
be
fair
to
say
we
could
move
to
beta.
But
at
this
point
on
skeptical,
okay,.
E
A
I
think
so
Wendy
you
kept
was
initially
proposed.
It
was
the
idea
was
to
propose
that
new
architecture
with
the
metadata
containment
set,
there's
one
calling
for
now
to
apply
for
all
providers,
with
the
assumption
that,
like
we
want
kubernetes
in
general
to
be
more
sensitive
to
API
quotas
and
I
push
back
a
little
bit
on
that,
because
it's
a
pretty
big
change
to
to
kind
of
force,
all
the
providers
to
adopt,
so
that
so
that's.
A
E
E
Needs
to
be
part
of
that
deadline.
If
the
work
they're
doing
is
out
of
tree
the
work
they're
doing
in
tree,
then
I
totally
I
think
it
matters.
But
it
also
seems
like
maybe
we're
breaking
this
into
two
caps,
which
I
think
is
a
very
good
idea
in
terms
of
tracking.
What's
in
tree
and
what's
out
of
tree.
I
E
I
E
I
A
Think
it
my
goal
also
worth
considering
like
a
higher
level
kept
that
addresses
API
quota
sensitivity
in
general,
not
like
so
like
pretty
much
like
going
through.
A
E
A
A
A
E
E
A
There
was
some
debate
around
how
we
want
the
the
GA
labels
for
volumes
to
be
applied
and
like
if
want
like,
do
we
want
like
when
we
do
an
upgrade?
Do
we
want,
like
both
the
G
and
the
beta
labels
to
be
applied,
and
then
what
happens
on
rollback
and
there's
a
lot
of
educators?
We
have
it
considered,
so
we
pushed
it
back,
but
I'm
hoping
to
get
back
to
it
for
117,
okay,.
E
E
E
A
E
E
Is
it
is
absolutely
and
I
mean
well,
as
as
Maria
mentioned,
it's
a
pretty
short
one.
Seventeen
is
a
pretty
short
cycle,
so
I.
Imagine
even
the
things
that
where
we
approve
like
the
beta
requirements,
there's
a
good
chance.
We
won't
actually
get
them
finished
just
because
it's
such
a
short
cycle,
so
yeah,
what's
at
least
set
ourselves
up
for
success
on
118.
E
E
A
Yeah
one
thing
that
that
is
probably
worth
mentioning,
which
is
a
kept
by
Signet
work,
but
it
directly
involves
us
because
this
is
only
a
cloud
provider
feature,
but
there
is
the
finalizar
support
for
a
serviceable
bouncer
which
went
beta
and
116
and
enabled
by
default,
and
then
in
117.
Probably
you
will
see
a
that,
so
just
extra
contest
for
folks
that
missed
it
essentially
without
finalized,
errs,
like
your
cue
controller
manager,
can
delete
a
service
resource
associated
with
a
little
balance
syrup.
If
and
somewhere.
A
If
that
errors,
then
you
end
up
deleting
the
service
resource,
but
not
to
being
the
load
balancer,
and
so
the
control
plane
does
know
to
clean
up
the
low
bounce
resource,
so
that
would
never
get
cleaned
up.
The
finalizer
ensures
that
the
service
resource
can't
be
deleted
until
all
of
its
associated
cloud
provider.
Resources
are
also
deleted,
so
it
just
ensures
that
you
can't
have
dangling
resources
on
your
cloud
provider.
A
A
Since
I
kind
of
did
most
the
work
already
I
just
need
to
sync
up
with
some
folks
in
six
storage
to
make
sure
we're
happy
with
the
way
were
applying
the
the
zone
and
region
Vall
the
zone
in
region
label
for
dynamic
PBS.
So
that
was
where
it
got
tricky.
Otherwise,
I
think
we're
good
to
work
for
with
this.
A
J
A
Yeah,
honestly,
at
this
point,
like
any
sort
of
Doc's,
would
be
great.
A
lot
of
providers.
Don't
have
Doc's
so,
like
I,
think
we're
just
trying
to
prioritize
like
let's
try
some
Doc's,
because
there's
a
lot
of
there's
a
whole
slew
of
things
that
aren't
documented,
for
example
like
annotations
for
low
balance
or
40lb
things
like
that
like
we
need
to
have
some
place
to
write
Docs
and
so
for
vSphere.
We
wrote
a
github
book
or
a
get
book
that
has
up
into
Doc's.
A
A
A
B
He's
the
right
person
to
assign
that
to
and
yeah
that's
that's
a
good
place
to
start.
You
know,
and
you
know
she
I,
don't
think
she's
officially
working
on
it
yet,
but
if
you
Kindle
Nelson
who
goes
by
Diablo
Rojo,
she
might
also
be
good
to
CC
on
it.
Just
to
make
sure
this
woman
at
the
foundation
sees
what's
going
on
with
the
requests,
so
they
can
prioritize
it.
A
A
B
A
A
A
Don't
remember
what
that
action
item
here
was
from
last
time.
I
think
it
was
just
we
just
so
I.
Think
AWS
is
the
only
cloud
provider
that
uses
the
cluster
ID
functionality
because
they
need
it
for
tagging
everything
the
cluster
creates,
so
I
think
the
action
item
was
if
any
of
the
provider
uses
the
cluster
ID
functionality
and
if
not,
if
we
can
propose
the
way
to
just
like
build
it
into
the
UW's
product
provider,
only
and
not
have
it
in
core.
A
I
think
p2
next
is
fine
to
me.
This.
A
A
Decoupling
hell
providers
from
the
kubernetes
end-to-end
framework,
so
I
am
involved
in
the
testing
comments,
a
project
insuk
testing
on
this
one.
It's
just
a
long-running
effort,
so
I
just
keep
it
in
the
next
milestone.
But
if
anyone
else
in
this
thing
is
interested
in
this
work,
this
work
is
mostly
involving
trying
to
get
rid
of
all
the
provider
specific
logic
in
the
end-to-end
testing
framework,
which
is
mostly
geared
towards
GCP
and
AWS,
and
so
we're
just
trying
to
make
it
more
provider
agnostic.
A
This
came
up
because
we
don't
have
a
good
way
to
so,
like
a
lot
of
cloud
providers
that
create
service
set
load
balancers,
they
have
more
than
one
resource
associated
with
the
load
balancer
like
public
ip's
and
firewall
rules
and
whatnot.
There
was
a
request
some
time
ago
to
track
these
resources
better,
but
I
put
the
finalizer
support.
I,
don't
think
we
need
to
do
that
anymore.
A
E
My
only
concern
there
would
be
that
the
the
finalizar,
whatever
code
you're
using
for
the
finalizer,
would
need
to
be
able
to
track
what
the
allocated
resources
are.
If
the
cloud
provider
has
something
to
do
that
automatically.
That's
fine
but
I,
wonder
if
and
maybe
it's
just
an
annotation,
but
if
there
may
need
to
be
some
sort
of
reference
to
the
cloud
provider,
specific
feed
gas
fields,
structures
that
need
to
be
cleaned
up.
Mm-Hmm.
J
That's
basically
what
we
did
at
digitalocean.
We
introduced
an
annotation
that
helped
us
keep
track
of
what
load
balance
is
associated
with
a
service,
because
we
could
run
into
the
case,
and
this
might
be
specific
to
our
cloud
where
users
would
do
things
like
change
the
load,
balancer
name,
and
then
we
would
start
losing
track
of
things
and
that's
and
that's
why
we
did
that
edit.
The
annotation.
H
A
Yeah,
oh
yeah
and
I
committed
on
this
for
the
meeting
I
had
mentioned.
We
should
just
like
I,
think
the
cleanest
solution.
Right
now
it
was
just
moving
CMD
clap
which
will
manager
to
kubernetes
/
clap
provider.
That
way
like
any
cloud
provider.
Implementation
really
just
has
to
import
the
cup
provider
repo
well
and
Plus.
You
know
client
go
and
whatnot,
but
it
should
be
a
pretty
clean
dependency.
Just
if
it's
just
a
cloud
provider
repo
but
I
wanted.
E
A
Okay
yeah,
so
my
perspective
was
mostly
go
dependency
related
because
I
know
the
biggest
pain
point
that
at
least
I
hear
is
having
to
vendor
of
kaykai.
So
the
hell
is
my
perspective
on
it,
but
yeah
I
agree,
there's
any
other
issues
they
should
be
raised,
but
I'd
like
to
prioritize
this
with
the
go
tent
good,
the
go
dependency
issue
in
mind
and
then
tackle
everything
as
a
secondary
issue.
If.
E
D
A
J
A
You
seen
it
awesome,
I'm
gonna
bump
this
to
p1
for
this
release,
just
because
I
think
if
you
have
good
idea
where
we
want
the
command
the
complementor
library2go,
I
think
we
should
do
it
soon
to
kind
of
avoid
people
trying
to
vendor
in
the
corker
bradshaw
story,
cool
support
and
submit
a
service.
So
this
is
the
what
we're
talking
about
earlier
with
Craig
I'm
gonna.
Keep
this
low
priority
for
now.
Until
we
have
that
discussion
again,
around
got
a
capacity
equipments.
F
Yeah
for
this
one,
there
was
some
discussions
with
sig
note.
So
basically,
the
idea
is
that
we
have
to
make
this
knob
change
feature.
Basically,
if
you
want
to
to
detach
volumes
when
I
know
this
shut
down,
the
user
have
to
explicitly
opt-in
for
this.
So
this
this
effort
would
span
across
at
least
two
releases,
so
I
I'm
expecting
to
work
on
this,
but
the
feature
would
not
ship
for
117.
F
J
E
E
J
E
A
A
J
So
I
think
for
my
point:
there
was
there's
some
uncertainties
about
certain
flag.
Well,
I
just
wasn't
sure
whether
those
were
still
needed
or
not
and
I
think
this
kind
of
got
lost
along
the
way.
So
if
somebody
could
just
help
me
better
understand
which
of
those
who
really
need
or
not
really
need,
but
just
need
kind
of
a
final
sign-off
and
confirmation
and
then
I
can
move
on
and
try
to
clear
up
anything,
that's
not,
but
as
far
as
I
understand
it,
it
should
be
a
minority
of
flags
that
really
would
go
away.
Okay,.
A
A
J
F
A
E
E
A
A
Don't
think
we
need
to
worry
about
this
in
the
immediate
future,
all
right
that
was
it
who
for
backlog
any
other
issues,
folks,
when
I
bring
up
all
right
what
up
time.
So
if
you
do
have
issues
you
want
prioritize
for
117,
please
open
an
issue
in
the
kubernetes
Club
butter,
repo
and
I.
Think
that's
in
the
cloud
provider
Channel
and
then
we
can
discuss
how
we
want
to
prioritize
it
and
then.