►
From YouTube: SIG Cloud Provider 2022-08-17
A
All
right
welcome
to
second
cloud
provider.
It's
wednesday
august
17th
go
ahead
and
get
started
with
triage,
so
got
two
issues.
Two
new
issues
today.
B
To
see
if
it
can
be
replicated
with
out
of
tree,
because
if
it's
the
entry
driver.
A
A
D
A
A
I
don't
quite
want
to
remove
the
triage,
yet
until
it's
a
little
bit
more
sort
of
taken
by
sig
storage,
so
I'm
gonna
loop,
someone
in.
C
A
C
D
A
You
know
that's
kind
of
like
that.
D
The
alternative
right
like
if
you
get
to
a
point
where
you
have
so
many
load,
balancers
and
nodes
in
your
cluster,
that
it
just
takes
long
enough
to
reconcile
them
all
and
go
through
these
things,
like
you
know,
unless
we
come
up
with
a
different
way
of
kind
of
processing
that
cube
or
there's
some
other
way
to
hold
these
things
for
an
extra
loop
or
something.
B
A
Okay,
so
yeah
was
changes
to
the
service
controller.
It
was
not
backboarded
or
actually,
I
guess
this.
A
Yeah,
I
don't
know
this
sounds
like
part
of
it
was
fixed,
but
oops
probably
was
fixed,
but
not
all
of
it.
So
I'll
just
comment
and
see
if
this
does
need
a
cap.
A
Ideally,
let's
all,
let's
all
read
on
our
own
time
and
then
maybe
we
can
discuss
it
next
week.
D
A
B
We
want
to
yeah,
I
guess:
do
we
want
to
encourage
alexander
to
open
the
cup
as
nick
is
pointing
out,
or
do
we
think
someone's
gonna
need
to
pair
with
him
on
that?
B
A
Yeah,
I
think
I
think
he
I
mean
my
reading
is
that
somebody
is
already
aware
and
working
on
a
cap,
and
maybe
I
missed
it
up
here.
A
Okay,
cool,
okay:
I
think
those
are
the
only
two
issues
sitting
in
triage
for
this
one
I'll,
also
I'll
leave
it
out
of
triage
just
so
that
we
take
a
look
at
it
in
two
weeks
and
follow
back
up
and
cool
yeah.
Let's
go
through
what's
up
project
updates.
If
everybody
wants
to
take
a
few
seconds,
if
you
have
an
update,
just
add
it
to
the
list.
B
B
Make
it
a
italicized
or
something,
because
I
keep
doing
that
on
these
too
yeah
I'll
go
since
I
I
put
one
in,
but
I'm
sure
other
people
have
their
thoughts
and
progress.
But
a
while
ago
we
had
gone
through
and
opened
a
bunch
of
those,
I
think
nick.
You
opened
them
right.
B
Those
very
helpful
issues
that
were
just
like
basically
go
fix
your
tests,
people
and
I
wanted
to
say
that
we
have
been
paying
attention
to
that
and
in
provider
azure
we
went
and
got
all,
but
one
of
our
failing
tests
are
currently
fixed,
so
yeah.
If
you
click
on
that
link
and
search
for
azure,
there
is
only
one
that
is
currently
failing.
A
B
Awesome,
I'm
just
just
a
reminder
to
all
of
us
to
pay
attention
to
these
family
tests
issues
that
nick
had
opened
for
us
because
it
is
yeah
for
azure.
If
you,
if
you
search
for
azure,
you
will
find
there
is
one
right
now
yep,
but
I
think
that
we
can
all,
I
think,
it'll
be
impossible
for
all
of
us
to
have
zero
failing
tests
forever.
But
I
bet
we
can
all
follow
what
nick
asked
us
to
do
and
open
those
go
pursue
some
of
your
failing
tests
and
I
bet
it'll
be
possible.
B
A
No,
that's
really
good
news
and
amazing
job
all
right.
Kishore.
C
Yeah,
so
for
the
load
balancer
controller,
we
release
v2
4.3.
It's
a
patch
release.
Miner
fixes
and
security
updates
in
the
last
patch
release.
A
Cool
all
right,
I
think
that's
everything
that
I
see
so
unless
anyone
else
has
anything
I'm
going
to
move
on
to
agenda.
B
Yeah,
I
just
put
a
couple
since
we
were
agenda
light,
but
obviously
we
don't
have
to
spend
a
ton
of
time
on
them,
but
there
was
one
that
our
team
did
go
through
and
put
a
couple
of
updates
on
this
particular.
If
you
look
at
that
particular
one,
that
is
one
of
those
it's
been
open
absolutely
forever
and
we're
trying
to
get
it
through,
and
I
wanted
to
see
if,
because
we,
if
you
jump
down
to
the
bottom,
we
did
just
get
some
suggest.
B
You
had
made
some
suggestions
very
helpfully,
which
I
appreciate
and
we
did
get
a
couple
of
those
in,
and
I
was
wondering
if
those
you
know
the
comments
or
the
the
changes
that
were
made
like
helped.
You
had
a
few.
You
know
questions
back
to
you.
Basically,
I
just
wanted
to
make
sure
we
keep
this
one
moving.
A
A
Yeah
this
one
I
mean,
is
a
little
hard
for
me
to
review
because
I'm
not
super
familiar
with
the
rev
controller,
but
it
just
seemed
like
there
was
a
configuration
setting
that
was
going
into
the
per
route
struct
and
didn't
quite
seem
like
the
right
place.
So
I
was
kind
of
just
asking
if,
if
it's
possible
to
put
that
in
the
like
yeah
the
rock
controller
structure
or
something
else,
that's
kind
of
more
global
or
just
a.
I
wonder.
B
E
I
took
a
cursory
look.
I
know
bridget
actually
pinged
me,
I
think
two
weeks
ago,
and
I
promised
that
last
that
two
weeks
ago
that
week
was
bad
and
I'd,
look
at
it
at
the
next
week
and
yeah.
I
think
I
needed
a
reminder
and
I
apologize.
I
will
try
to
get
it
to
it
later
this
week.
I've
actually
got
it
written
on
a
post-it
note
now
and
on
my
laptop.
B
A
Yeah,
no
I'll
take
it
I'll.
Take
another
look
and
see
did
do
you
know
if
the
author
made
a
change
to
the
setting
at
all,
I
don't
or
were
they
just
responding.
B
B
A
Doing
the
right
thing
totally
makes
sense
cool
yeah
I
will
get.
I
will
give
it
another
review.
B
B
Oh
yeah,
so
I
was
looking
at
this
one
too.
This
is
actually
a
brand
new
one
and
it's
just
a
bug
fix
and
it's
like
oh
cool,
like
maybe
this
is
a
issue
that
we're
actually
going
to
get
fixed
without
it
taking
forever.
B
I
think
so
this
is
it
fixes
that
particular
issue.
So
if
you
click
through
to
the
issue
that
it
fixes,
this
is
one
of
those
you
know
we
we
have
this
one
in
area
provider
azure
and
we
did
actually
triage
it
here
and
get
this
one
like
hey.
We
should
probably
fix
this
and
it
looks
like
it
was.
A
B
A
Right
right,
yeah,
you
need
the
lgtm
so
yeah.
I
will
just
take
a
really
quick
look
at
it
and
yeah
give
the
lgtm
mmsm.
B
E
E
So
there
had
been,
there
had
been
a
person
who
had
been
requesting
the
ability
to
run
extra
cloud
provider
specific
logic
in
the
ccm
without
having
to
worry
about
doing.
You
know
basically
have
a
controller
that
kicked
in
after
so
the
cloud
like
the
cloud
life
cycle
controller,
and
so
one
of
the
suggestions
outstanding
that
I
had
made
had
been
pos
had
been
on
top
of
some
work.
E
That
nick
was
doing
where
nick
was
basically,
you
know,
setting
whether
using
the
cloud
provider
equals
external
to
indicate
whether
or
not
to
add
the
cloud
tank
which
overloads
it
a
little
bit.
And
so
one
of
my
counter
suggestions
had
been
given
that
there
was
another
customer
who
wanted
to
be
able
to
run
extra
controllers.
E
Is
that
maybe
what
we
would
like
to
do,
rather
than
the
cloud
provider
external
setting?
The
taint
is
actually
have
a
taint
list
for
new
nodes
that
most
people
would
just
set,
if
they're
a
cloud
provider
to
just
being
the
the
normal
cloud
tank.
E
But
if
you
had
extra
controllers
you
wanted
to
run,
you
could
add
them,
you
could
add
extra
taints
and
then
check
that
they
got
cleared.
So
that
makes
I'm
not
sure
if
I'm
elaborating
well
on
that,
but
did
that
make
sense
to
folks.
A
Yeah
I
mean
I
follow
it
because
I
we
had
that
conversation.
Do
you
happen
to
have
the
link
to
like
the
original
issue.
E
E
Yeah,
it's
not
exactly,
and
that
was
just
privately
slacked
to
me,
but
I
thought
it
was
an
interesting
enough
thing
to
bring
up.
A
D
E
I
I
personally
would
advocate
that
this
doesn't
require
a
full
cap,
because
all
we're
doing
is
adding
a
flag,
but
I
do
think
it
should
at
minimum
be
an
issue.
That's
a
very
good
point,
and
when
push
comes
to
shove,
if
either
nick
or
andrew
says
they
think
it
should
be
a
cap,
I
will
defer
to
them.
E
D
A
Yeah
yeah,
I
think
we
should.
I
have
a
feeling
it
will
turn.
I
mean
if
it's
like
a
really
short
cup.
I
I
wouldn't
be
surprised
if
it
ended
up
being
one
just
because
it's
like
a
api
change
kind
of
thing,
but
but
this
is
the
this
is
the
pr
that
walter
was
referring
to
I'm
basically
just
changing
sort
of
deprecating
the
cloud
provider
flag,
changing
it
to
a
boolean
configuration
where
it's
just
external
or
not
external
and
external,
directly
means
apply
the
taint.
A
So
that's
where
the
like
question
of
okay,
if
we're,
if
we're
changing
that-
and
we
also
have
this
idea
of
people
wanting
to
be
able
to
put
their
own
paints
on
nodes
and
then
have
their
own-
you
know
controller
deal
with
that.
Then,
should
we
combine
these
into
one
thing.
A
E
Would
it
be
sufficient
if
I
filed
the
issue
and
just
ceced
the
person,
I
think
so
yeah
all
right?
I
will.
I
will
get
an
issue
filed.
D
E
So
I
I
so
maybe
it
does
so
my
thought
would
be
that
if
you
had
that,
what
we're
saying
is,
if
you
have
cloud
provider
external
it'll,
add
the
x
until
cloud
provider
goes
away,
it'll,
add
the
uninitialized
taint
and
that,
but
we
add
verbiage
to
that,
almost
like
what
nick
has
here
that
says,
you
know
going
forward,
there's
a
cloud
node
taint
and
you
would
be
expected
for
most
of
us,
google,
azure,
etc.
You'd
be
expected
to
say
cloud.
Node
taint
equals
uninitialized
for
cops
and
other
ones.
E
D
I
feel
like
that
will
need
a
cup
because,
like
right
now,
the
expectation
is
that
if
I
set
cloud
provider
external,
I
get
the
uninitialized
team
and
I
would
expect
that
in
the
future,
when,
if
the
cloud
provider
flag
goes
away
and
everything
has
to
be
external
cloud
providers,
then
shouldn't
the
kubelet
like
always
come
up
with
that
up
like
having
to
switch
to
apply
it
myself
is
going
to
be
a
big
change.
I
think,
for
a
lot
of
there's.
E
Something
to
that
I
mean
the
problem
is
that
today,
if
you
just
don't
provide
the
cloud
for
the
cloud
provider
equals
external
flag,
it
doesn't
set
the
uninitialized
taint
and
our
testing
and
cops
and
a
lot
of
the
single
machine
setups
require
that
paint
not
be
set.
D
E
A
Okay
yeah,
so
I
would
say,
poulter
file
the
issue
that
represents
that
point
of
view,
and
then
we
can
discuss
on
the
issue
whether
we
need
a
cup
or
not.