►
From YouTube: SIG Cloud Provider 2020-04-15
Description
1.19 backlog grooming
A
Hi
everyone:
this
is
the
Sig
cloud
provider
meeting
on
Wednesday,
April,
15,
20
20.
Please
remember
the
CN
CF
code
of
conduct
and
be
nice
to
everyone,
especially
in
these
slightly
more
stressful
times,
and
with
that
I
will
hand
it
to
my
co-host
Andrew
to
start
with
the
first
agenda
item,
which
is
the
credential
provider
cap.
B
B
We
want
to
simplify
the
way
a
provider
implements
the
external
registry
pretender
provider
so
that,
instead
of
the
API
requiring
the
instead
of
the
you,
the
consumer
vai
having
to
know
what
the
executable
commands
and
arguments
are,
we're
just
standardizing
what
the
arguments
are
and
what
the
commands
are
going
to
be
for
the
for
the
external
binaries
of
culet
runs,
and
so
the
provider
just
have
to
implement
every
command
that
the
plug-in
supports.
So
the
first
plug-in
that
is
win
support
is.
B
B
Name
of
the
credential
provider
and
then
the
first
argument
being
get
credentials
and
then
there's
going
to
be
some
some
other
fields
that
indicate
like
extra
agreements
and
whatnot
so
just
wanted
to
bring
this
kept
update
to
the
sake
to
get
broader
visibility
and
and
yeah.
Hopefully,
you
can
get
this
kept
approved
early
on
so
that
it
can
start
work
on
it
early
in
the
cycle
just
for
context
for
118.
A
D
D
A
A
E
A
You
very
much
Google,
it
looks
like
we
should
actually
make
the
180
have
the
118
alpha.
Also,
interestingly,
it
looks
like
we've
got
at
least
one
outside
set
of
contributors
who
were
actually
also
working
on
getting
CCM
working
on
GCP,
so
I'm
gonna
reach
out
to
those
as
well
Andrew
Chris.
Do
you
have
an
update
on
VMware.
B
B
Supporting
a
Yambol
version
of
the
clapping
fake
I
know,
most
providers
use
and
I
hope
we
one
else
am:
oh
that's
so
both
of
those
are
alpha,
but
otherwise,
like
no
I
mean
our
new
provider
is
GA,
and
so
we
know
we
already
have
products
kind
of
shipping
with
the
other
two
provider,
which
is
really
good
and
we
have
users
are
using
it.
The
only
thing
left
on
our
plate
is
finishing
the
work
to
remove
the
Daintree
one.
That's
about
it.
Also.
B
A
B
A
A
B
Yeah
sure
I
mean
I
can
be
that
one
to
share
this
show
the
screen,
but
the
control
together,
so
I
just
yeah
pretty
much
just
opened
all
the
tickets.
Do
we
want
to
go
from
oldest
to
newest?
Yes,
let's
do
that?
Yes,
that
sounds
reasonable.
Alright,
oh
that
reminds
me.
I
need
to
first
create
a
new
milestone.
A
B
Yeah
so
yeah
I'm
gonna,
so
we
I
think
that
a
few
years
ago
there
was
some
efforts
to
improve
the
documentation
on
a
per
provider
basis
and
I
know.
Some
providers,
like
meteor
and
I
sure,
have
worked
on
some
Doc's.
I
know
the
folks
that
in
WS
haven't
so
I
think
what
we've
been
doing
for
pretty
much
most
releases
have
been
like.
This
is
still
be
zero.
Folks,
you
want
to
promote
folks
to
work
on
docks,
so
we
just
leave
this,
as
is
milestone.
A
I'm
actually
starting
to
get
some
resources
so
I'm
actually
hoping
that
in
119
we
can
make
something
happen,
especially
since
we
should
actually
have
a
reference
implementation
I'd
like
to
get
the
docks
in
place
for
that
and
for
gke,
we're
gonna,
add
lifecycle
frozen
after
that
looks
like
we
already
did.
Yeah
can.
B
B
B
Okay,
this
is
related
to
the
big
package
that
I
don't
think
what
you
do
is
right
now,
but
I
know
some
of
my
previous
co-workers
at
digitalocean
wanted
to
work
on
this
and
they
kind
of
worked
around
it
by
supporting
an
annotation
that
flips
the
load
balancers
name,
so
they
kind
of
did
their
own
thing
for
this
I'm.
Just
going
to
keep
this
on
next
with
p3
like
that,
I
think
this
is
one
of
those
things
were
for
service
set
little
bouncer
I.
B
A
B
One
yeah
the
KCM
to
CCM
migration,
so
the
cap
is
in
I
have
some
work
to
do
to
find
a
home
for
the
migration
or
the
config
that
config
types
were
using
for
the
to
to
manage
the
migration
of
the
controller's.
Yes,
yeah.
I
think
this
is
just
on
me
and
Walter.
Just
in
combat,
so
I'll
continue
p0
for
119
yeah.
A
B
B
Usage
requirements,
the
cluster
ID
yeah
I
think
this
I
think
P
2
at
next
make
sense.
I
think
this
is
on
someone
from
Amazon
to
work
on,
because
cluster
ID,
as
far
as
I
know,
is
only
used
by
the
Amazon
provider,
so
muscle
next
P
to
make
sense
and
then
we'll
hopefully,
Nick
will
get
to
it.
But
if
he
doesn't.
B
A
D
A
D
B
B
B
But
yeah,
like
I,
had
previously
put
this
at
a
pretty
low
priority,
because
there
are
ways
to
there
is
ways
to
optimize
this
in
the
cloud
provider
like
there's
no
reason
to
change
no
controller
to
do
this,
like
you
can
have
an
internal
cache
and
your
provider
for
this.
You
don't
want
to
repeatedly
making
the
calls
or
you
can
change
your
cloud
provider
to
have
higher
kotas,
but
yeah
I
think
this
is
low
priority.
B
Maybe
maybe
for
Rock
controller,
there
are
some
optimizations
we
could
do
because
rel
controller
is
pretty
aggressive
like
it.
It
lists
all
routes
on
every
sync,
so
there
may
be
a
way
there
actually
might
be
something
we
can
do
but
again
like.
If
cloud
providers
aren't
it
does
not
like
it
doesn't
seem
like
it's.
It's
breaking
scalability
tests
or
anything
like
it
seems
like
a
barrier
provider
specific
problem.
They
saw
the
quarters
on
provider.
Side
agree
all.
A
B
B
Nice
cool
decoupling,
copper
virus
from
the
kubernetes
intern
framework,
so
George,
who
runs
the
testing
common
sub
project.
He
has
a
cap.
Actually
he
has
a
draft
kept
open
right
now
for
doing
this
work.
So
sorry
actually
clarify
not
specifically
for
decoupling
the
cloud
providers
that
he
has
a
proposal
open
for
staging
the
testing
framework
and
I
have
it
on
my
plate
to
remove
the
provider
framework
inside
the
providers
inside
the
testing
framework
and
then
have
it
call
out
to
each
individual
talk
about
agree
before
so.
I
think
119
makes
sense.
B
Okay,
yeah
so
I.
Remember
the
context
for
this
so
Timo
that
works
on
at
digitalocean,
with
like
a
way
so
like
for
some
providers.
When
you
create
a
little
balancer,
you
don't
just
create
a
little
balancer
on
the
cloud
brother.
You
also
create
a
firewall
and
the
may
be
a
separate
resource
for
the
public
IP
assigned
to
the
load,
balancer
and
so
I
think
there
was
some
conversations
around.
B
B
C
B
C
B
D
C
E
B
B
Cool
all
right,
handle
volume
schedule
even
knows
oh
shut
down,
so
it
seemed
had
a
bunch
of
context
on
this,
so
I'm
gonna
I'll
actually
enclose
this,
because
I'm
not
sure
these
teams
working
on
this
anymore
and
the
cloud
provider
is
just
one
part
of
it,
but
I'm
pretty
sure
this
is
a
signal
issue
at
a
high
level.
So
if
we
want
to
get
this
done,
we
should
just
open
issue
in
sig
note.
B
B
B
A
A
B
Think
it
might
also
help
I
think,
maybe
so
I
think
maybe
like
for
119.
You
should
just
focus
on
like
scoping
the
work,
because
every
time
this
topic
comes
up,
we
don't
really
get
anywhere
because
the
work
there's
like
so
much
work
to
do
in
this
area
and
I
think
most
folks
kind
of
get
scared
away
by
the
amount
of
work
involved.
So
maybe,
if
we
spend
some
time
just
like
defining
the
scope
of
this
work
and
then
breaking
down
to
smaller
chunks,
then
we
can
create
separate
issues
for
a
chunk.
Sounds.
A
B
A
Tagging
the
provider
I
get
but
I
mean
just
looking
at
the
API
above
and
I.
Think
I
just
need
to
understand
better.
This
method
is
implemented
by
a
provider,
so
I'm
trying
to
understand
why
you
need
to
pass
in
a
provider
string
and
what
happens
if
the
provider
string
doesn't
match
the
provider
implementation.
B
A
C
All
right,
Steve,
Wong
joined
late
because
I
had
a
meeting
in
front
of
us
go
over
I'm
with
the
where
I
was
co-chair
of
the
VMware
sake
and
I'm
kind
of
a
lurker
I'm
the
we're
proud
provider
keeping
tabs
at.
What's
going
on
because
I'm
co-chair
of
the
VMware
children
at
each
user
group.
So
looking
out
for
people
running
kubernetes
on
top
of
the
PM
with
platform
cool.
C
A
A
One
quick
mention
for
everyone:
I
should
have
mentioned
it
at
the
top.
If
you
go
into
your
participants
list,
there
are
a
series
of
flags
that
you
can
add.
So
if,
at
any
point,
you're
in
a
any
kubernetes
cig
meeting-
and
you
need
to
do
something
like
you
want
to
raise
your
hand
to
indicate
I
will
make
an
example.