►
From YouTube: kubeadm office hours 2019-08-28
A
A
A
A
A
A
A
So
it
has
a
blocking
operation
for
a
number
of
occasions
in
the
cake
here
poster
and
one
of
the
occasions
PBM,
it's
a
pretty
minor
change,
it's
something
that
is
required
for
the
new
HD
version,
but
this
is
moved
on
approval.
I
also
think
that
we
don't
have
an
image
I'm,
pretty
sure
we
don't
have
an
image
for
the
new
HD
version,
either
GCR.
A
Another
one
is
that
I
saw
I
saw
some
movement
in
a
Windows
work
for
116.
This
is
the
TR
for
the
new
Windows
repository.
It's
basically
windows
doors.
This
is
where
the
documentation,
basically
not
communities
website,
is
going
to
link
to,
and
this
is
what
would
basically
instruct
the
users
to
use
a
particular
powershell
script
or
a
set
of
powershell
script.
I
think
this
should
be
embedded
for
116.
A
Maybe
I
should
add
the
quality.
So
the
story
there
is
that
users
bow
on
PowerShell
scripts,
they
arunda
locally
and
supposedly
they're
going.
So
you
know
the
cube
a
diem
and
all
the
specifics,
like
you
know,
two
pods
in
Windows
deserves
run
us
a
demon
sect,
so
it
downloads
a
binary
compound
from
windows
and
also
it
setups
of
a
specific
version
of
flannel,
which
is
also
specific
to
Windows.
It
runs
as
a
service
instead
of
again
at
the
USA,
so
I'm,
just
noting
that
maybe
also
is
going
to
91
16.
A
Hopefully
like
the
final,
he
says,
I
created
doc
document
placeholders
for,
like
all
the
documentation,
we
talk
difference.
These
images
need
to
find
the
time
to
work
on
these.
These
are
still
placeholders
and
the
deadline
is
the
third
of
September.
So
that's
all
the
pieces.
Does
anybody
have
any
topics
you
want
to
discuss.
B
B
Today,
we
with
ability
was
on
whether
we
should
actually
even
deprecated
and
eventually
drop,
specifying
full-blown
component
config
on
I,
guess
part
of
the
config
file
during
Q
beta
a
minute,
and
instead
to
replace
this
with
customized
and
basically
allow
users
to
just
go
and
specify
several
patches
or
like
patch
set,
which
can
be
applied
to
component
config.
So
this
patch
set
can
in
theory,
then
we
use
the
game
during
upgrade
over
a
newly
generated
one
by
the
by
cube
ATM,
and
if
the
past
does
not
apply.
B
B
Basically,
keep
up
with
the
rules
of
capi
machinery,
but
like
right
now,
I'm
deliberating
on
this
and
scratching
my
head
and
dumping.
Some
ideas
into
a
document
and
I
I
hope
that
by
the
next
week,
I
can
actually
put
up
a
little
bit
more
structure
to
it
and
like
dump
it
in
the
open
and
see
what
what
comments
you
have.
A
A
B
So
right
now
it's
going
to
be
over
cube,
ATM
generated
defaults,
at
least
that's
what
I'm
thinking
of.
But
it's
still
like.
It's
still
not
clear
to
me
whether
this
should
be
done.
For
example,
during
cube
ATM
in
it
and
cube
ATM
upgrade
apply
and
basically
allowed
the
budget
component
of
it
to
be
stored
in
the
coffee,
Maps
and
then
fetched
over
into
joining
and
a
great
notes
and
gather
ways
to
basically
do
as
we
are
actually
doing
right
now
with
customized
and
basically
allow
users
to
do
a
Burnout
component
config
patching.
B
But
this
is
of
course
going
to
be
applied
only
if
we
actually
narrow
down
the
component
config
component
configure
that
we
actually
managed
to
only
those
that
are
start
like
only
those
that
are
applicable
to
static
pots
and
the
cubelet.
So
basically
we're
going
to
avoid
to
proxy
and
probably
queue
proxy
is
going
to
be
managed
in
the
future
by
a
by
another
operator,
I.
B
Well,
if
we
are
going
to
be
like
going
to
no
local,
it's
going
to
be
hard
for
queue
proxy.
So
probably
we
should.
We
can
actually
implement
that
with
some
in
its
container
but
like
what
that's
not
the
point
of
this.
So
and
probably
it's
going
to
be
like
taken
over
by
another
operator
in
the
future,
so.
C
B
As
the
component
config
llamó,
so
basically
you
can.
You
can
actually
specify
right
now
in
a
config
file
to
Q
beta,
aiming
it
not
on
the
cube
ATM
types,
but
you
can
actually
specify
currently
a
couple
of
different
types.
So
those
are
the
supported
ones.
Right
now
are
the
cubelet
component
configuration
and
the
cube
proxy
component
configuration
so
the
cube
root,
one
is
I,
think
cubelet,
dot,
config,
dot,
Kate's
dot
IO
and
it's
a
V
1
beta
1.
D
Yet
Ross
just
wanted
to
say
thank
you
for
working
on
this
I
didn't
know
that
you
had
taken
this
item
and
if
you
want
to
just
talk
design
and
UX
more,
please
feel
free
to
ping
me
outside
of
the
meeting
yeah.
Thank
you,
because
I
can
definitely
think
of
some
ways
that
we
can
support
users
as
they're
working
with
different
versions
of
configs
when
they
are
owned
by
the
binaries
yeah.
Thank
you.
E
E
E
A
You
can't
appear
you
just
have
it
here
today.
Ok
I
want
to
try
to
quickly
say
about
the
important
topic,
because
so
recently
I
found
that
we
import
both
completely
ignores
underscore
test,
not
go
file,
so
it
completely
deserved
one
from
test
files.
So
you
came
import.
You
know
the
doctor
is
the
key.
You
know
this
file
and
import
boss
is
not
going
to
questions
so
we
considered
disabled
Indians,
gender
PR
to
fix
it
and
he
basically
his
PR
extended.
A
You
know
import
boss
to
support
these
files,
but
then
we
found
out
that
it
also
the
test
files
can
pull
a
lot
of
indirect
dependencies
from
other
from
test
famous
interpreter
and
Jordan
commented
that
this
is
not
ideal.
We
should
only
have
one
level
of
imports.
You
know
checking
one
level
of
imports.
Investments
and
indirect
dependency
should
not
be
in
fact,
so
there
was
no
better
middle
ground
and
we
decided
to
completely
again
stop
tracking
testing
balls
but
dismissed
sorry.
A
A
My
I
think
that
the
biggest
offenders
here
are
the
packaging
products,
see
the
packaging
couplet.
If
we,
you
know,
if
we
remove
these
I
think
that
all
of
indirect
dependency
is
going
to
be
gone,
are
going
to
be
gone,
but
I
just
wanted
to
mention
that
about
import
most
also,
it's
kind
of
buggy
and
I
found
some
weird
bugs
when
we
were
merging
some
hd/dvrs.
It's
you
know,
select
regular
expressions
combined
with
our
prefixes
result
in
some
weird
stuff,
so
yeah
I
just
wanted
to
add
this.
Miss
comment
for
purpose.
D
I'll
just
take
this
chance
to
PSA
a
little
bit
about
what
I'm
thinking
for
addons
work.
We've
been
looking
at
the
code
on
how
we
can
put
it
in
two
phases
and
the
the
logic
is
seeming
to
be
pretty
good.
When
there's
an
add-on,
installer
configuration
supplied
in
the
Kuban
m
config
and
the
feature
gate
is
enabled.
D
Well
when
the
feature
gate
is
enabled
we'll
will
use
the
default
add-on,
installer
configuration
when
an
add-on,
installer
configuration
is
provided,
the
user
will
be
overriding
whatever
the
default
is
and
the
way
we
can
pass
in
values
like
strategically
from
Kubb
idiom
for
defaults.
Programmatically
is
probably
going
to
be
through
a
JSON
patch
list
that
is
supported
by
the
API
and
then
the
Installer
runtime
can
construct
a
just-in-time
customized
file
that
references.
The
network,
resources
that
are
needed
while
still
supplying
the
patch
file,
so
yeah,
just
kind
of
where'd.
D
There's
no
capped
at
the
moment
so
as
far
as
making
116
that
sounds
pretty
unreasonable
from
kubernetes
like
release
team
perspective,
but
I
know
we're
also
pretty
minion
on
alpha
features,
if
not
I
have
really
no
reservations
as
well
and
just
waiting
for
117.
So,
however,
you
all
feel
as
a
team
about
that
is
fine
with
me.
A
D
A
So
proposing
that
we
can
have
the
patches
on
the
side
of
cube,
ATM
and
once
cube
ATM
malls,
so
people
in
custom.
Oh,
what
is
the
structure
of
the
basic
component?
Config
is
patching
for
a
certain
atom
so
doing
require
forgiving.
It
downloads
like
this
default,
its
corporate
consent
from
the
other,
so.
D
There's
two
stories:
honest
one
is
that
an
add-on
operator
has
full
access
to
the
community's
API
and
are
back
and
from
that
perspective,
there's
really
no
code
that
Kubb
ATM
needs
to
run.
It
just
needs
to
say,
hey,
install
the
operator
and
let
it
go
do
its
thing,
except
that
it
needs
to
tell
the
operator
like
what
to
install
right.
So
if
I
have
cord
EMS
operator,
then
kuba
TM
is
probably
gonna,
be
at
some
point
responsible
for
after
the
core
DNS
operator
is
up
applying
the
core
DNS
resource.
D
D
That
then
references
a
git
repository
and
that's
like
you
know
you
can
muck
around
and
create
files
and
do
anything
you
want,
but
I
just
thought
it
would
be
a
good
API
Edition
so
that
when
you
are
actually
using
the
add-on
installer,
if
you're
trying
to
do
something
trivial
that
you're
not
having
to
make
a
tempter
and
then
write
a
customized
file,
that's
compliant
with
that
API
version
and
have
to
vendor
all
that
code
in
and
that
kind
of
thing.
So
it's
just
kind
of
where
my
head
is
at.
A
Say
so,
instead
of
like
currently
cue
a
cube
ADM
like
for
cube
proxy,
the
user,
is
we
had
a
discussion
about
corporate
config
by
the
way
before
he
joined
you
did
watch
the
recording.
I
guess
mostly
was
explaining
his
going
work.
So
if
key
proxy
me,
we
currently
all
keep
boxing
on
the
Cuban
side
to
be
like
a
full-blown
component
config
that
is
piped
from
the
user
to
the
actual
atom.
So
I
guess
the
replacement,
for
that
has
to
be
something
that
is
on
a
similar
level
of
UX
simplicity.
So
maybe
we
have
to.
A
D
I
agree
and
I
think
that
there's
room
as
well
to
support
the
existing
UX
like
say
the
feature
graduates
from
you
know
alpha
through
to
GA
that
we
could
use
the
add-on,
installer
machinery
and
still
support
like
if
a
user
passes
a
cou
proxy
configuration
right,
then
it'll
create
a
default
add-on.
Installer
configuration
that
patches
in
there
good
proxy
config
and
that's
just
transparent
to
the
user.
Behind
the
scenes.
D
F
A
D
D
But
it's
more
of
a
niche
for
sure
to
be
supporting
like
the
cilium
could
route
or
use
case,
but
I
think
that's
the
whole
point
of
the
add-on
installer
is
that
it
really
it's
removing
some
of
these
operational
opinions
that
have
been
called
core
and
is
opening
up
the
playing
field
for
anybody
to
be
using
crouton
use
the
way
they
are
supposed
to
or
the
way
that
they
want
to.
Rather.
C
C
So
my
question
would
be
instead
of
trying
to
two
of
these,
these
patches
on
the
local
file
system,
but
it
makes
sense
if
we
talking
about
not
using
the
operator,
because
when
we
use
the
operator
the
printer
will
take
care
of
this,
but
bully
makes
sense.
If,
when
we
do
cube
ATM
in
it,
we
basically
upload
all
the
parties
that
we
apply
that
were
in
the
file
system
to
some
config
map
or
do
some
secret
or
something
similar
on
the
cluster.
C
So
then
the
UX
about
upgrading
would
be
release
in
the
sense
that
when
you
are
doing
QED
I'm
afraid
you
know
you
don't
really
need
to
have
these
parties
on
the
file
system,
but
they
will
be
downloaded
from
the
cluster.
So
we
have
the
base
the
new
base.
We
apply
them
on
top
and
then
we
apply
them
to
the
cluster
afterwards.
Would
that
make
sense-
or
you
know.
D
I
and
listening
to
you,
talk
really
makes
me
realize
some
things
about
this
UX.
That
I
think
we
could
make
more
helpful
we've
chatted
before
about,
like
the
state
store
model
of
kabillion
and
how
we
would
it's
been
mentioned,
that
it
could
be
desirable
to
deprecate
those
kinds
of
behaviors
where
we
are
automatically
and
transparently
from
the
user.
Just
getting
configs
from
magically
named
config
maps
in
the
cluster
and
then
using
that
to
persist.
D
Well,
I
I
think
it's
very
useful
like
to
a
cluster
operator
that
has
not
installed
the
the
the
like
initial
configuration
of
Covidien
to
be
able
to
see
a
history
of
what
was
the
intent
when
the
cluster
was
created
or
upgraded
in
the
cluster.
I.
Think
that
that's
a
good
move
to
like
record
history
and
then
also
fetching
that
history
back
from
the
cluster
locally
to
the
person
that
wants
to
operate
on
the
cluster
with
coop
in
iam
also
makes
sense
to
me.
But
I.
D
That
makes
a
lot
of
sense
to
me,
but
when
you
are
automatically
moving
the
source
of
truth
from
an
it
time
from
the
user's
laptop
into
the
cluster
and
then
when
upgrades
happen,
how
does
a
user
fork?
What's
in
the
cluster
right
when
they
need
to
change
the
version
of
calico
that
was
installed
or
something
to
eat,
compliant
with
new
API
versions,
that
is
I.
D
The
only
solution
that
sounds
reasonable
to
me
as
like
an
infrastructure
operator,
is
that
you
would
actually
fetch
those
things
from
the
cluster
or
from
the
authoritative
git,
repo
right
and
I
would
I
would
go
to
the
git
repo
first,
but
a
user
might
not
always
have
that.
So
that's
that's
pretty
dissonant.
From
what
you're
saying,
hopefully
I'm
hearing
you
correctly
and
you're
hearing
where
I
agree
and
disagree.
C
B
Think
that
it
has
its
ups
and
downs,
but
we
need
to
actually
deliberate
more
on
it.
It's
more
like
a
UX
thing
to
me
at
least,
but
it
does
have
some
has
to
like
probably
stored
up
the
patches
in
the
question
just
for
the
record,
keeping
but
at
least
automatic
patching
during
cup
grade,
especially
on
on
a
changed
version
of
the
component.
Config
is
not
really
that
desirable,
simply
because
we
may
end
up
with
with
some
weird
situational,
like
some
config,
that's
having.
B
Marching
field
names,
but
two
different
default
values
and
stuff
like
that,
so
we
may
end
up
in
some
weird
situations,
depending
on
the
changes
that
convicts
actually
introduce.
So
basically
it's
going
to
be
like
nitpicking,
and
probably
we
can
do
this
automatically
when,
whenever
the
component
configuration
is
not
changed
and
my
version
I
mean
that
the
API
version
of
the
like
diversion
of
the
config
binder
itself.
B
A
D
A
A
A
A
So
I
wanted
to
quickly
go
over
the
new
tickets.
You
know
not
essentially
not
necessarily
in
the
mouse
on.
There
are
a
couple
of
interesting
ones
here,
so
this
is.
This
is
a
very
complicated
one.
Basically,
someone
said
that
we
forgot
to
summon
the
Keyblade
coughs.
Why
are
we
right?
We
do
not
leverage
the
different
notation
on
the
table
and
this
is
only
the
primary
control
in
note.
A
If
you
join
more
nodes,
you're
going
to
see
this
symbolic
link
as
part
of
the
clan
certificate
instrument
coolant
off,
but
on
the
primary
node,
we
don't
do
this
and
the
users
have
to
manually
edit
the
completed
conference
if
they
want
to
memorize
notation.
So
we
that
kind
of
discuss
this
I
was
like
proposing
different
options.
A
Unfortunately,
one
of
them
a
good
because
we
are
going
to
break
existing
users
if
we
phases
around
or
if
we
start
making
assumptions
whether
certificate
rotation
is
enabled
or
not.
So
this
one
I
think
it's
pretty
much
on
top
of
the
list,
you
can
have
a
look
at
the
ticket
and
provide
some
feedback
on
it.
I
did
this
116,
but
after
my
initiation
I
don't
think
we
can
action
anything
out
of
it
for
this
cycle,
so
this
is
a
ticket
about
sorry.
A
A
A
Yeah
so
I
mean
I
was
I
was
thinking
between
another
three
years.
You
know
where
I
remove
items,
because
this
is
cleanup.
This
is
not
future
again
still
in
the
well.
The
contributor
has
signed
a
consent
from
the
cycles,
but
soon
I'm
going
to
go
over
all
the
116
kind,
design
kind,
future
and
I'm
going
to
move
into
a
cycle
because
we
are
approaching
cold,
sorry
feature,
freeze,
I,
think
it's
the
end
of
the
month.
Let
me
check
I
think.
A
B
So
basically,
this
is
a
pace,
library
and
it
actually
has
only
three
calls
to
K
walk
and
a
couple
of
those
are
refs
and
one
is
info
F.
But
the
info
Co
is
based
clear,
at
least
in
my
opinion,
and
these
calls
can
actually
be
translated
to
just
rapping
carers
and
returning
the
proper
wrapped
errors
back
to
the
core.
A
Yeah
Mike
contacted
me
on
private
and
I
proposed
the
solution
to
do
to
pass
an
interface
which
is
kind
of
more
entitled.
Like
you're
saying
we
don't
really
need
that.
Maybe
the
interface
basically
was
the
user
to
create
a
structure
that
has
methods
that
comply
with
the
demands
from
K
walk
in
for
stuff
added,
but
I
think
your
proposal
is
better.
The
only
the
only
downside
of
this
approach
with
the
returning
errors
is
that
the
diff
of
this
pair
is
going
to
be
bigger.
A
A
Actually,
we
haven't
merged
a
fearful,
and
so
during
join.
We
fetch
the
the
key
proxy
configured
from
the
question
and
if
the
user
skip
the
key
proxy
add
on
face
from
in
it,
the
key
prosecuted
map
does
not
exist,
so
we
have
an
error
during
join
and
the
idea
was
my
idea
was
to
not
error
out
in
the
kiosk
of
our
piece
missing
and
simply
populate.
A
A
So
we
fetched
by
mid
configuration
the
cluster
configuration
we
stitch
them
together,
which
I
don't
really
like,
but
you
also
take
the
proxy
and
we
sketch
the
proxy
inside
the
the
tip
of
Sentinel
type
inside
the
basically
the
internal
types
for
keeping
em
as
component
coffee
and
instead
of
fetching
this
from
the
cluster
I
was
thinking.
Maybe
we
can
simply
create
the
defaulted
to
proxy
type
or
empty
empty
structure.
I,
don't
know
I,
just
don't
want
to
error
out.
Thank
You
proxy
as
a
config
bub
is
missing.
A
D
D
B
I'm
not
sure
actually,
but
we
can.
We
can
just
ignore
the
error
and
the
cap
box.
Configuration
is
then
simply
going
to
be
like
defaulted
during
the
dynamic
defaulting
of
the
whole
configuration
so
the
next
time
we
do
actually
upload.
It's
probably
going
to
be
uploaded,
but
I
can't
remember
actually
the
cult
that
process
so
in
any
way.
We
should
probably
try
to
just
disable
the
like
tearing
out
and
simply
treat
that
as
an
empty
config,
which
is
going
to
be
a
default
to
Derek,
Derek,
coffee
and.
A
Right
I
want
to
take
action
on
this
able
to
see
it
appeared
like
today
or
tomorrow
about
this
I'm
going
to
assign
reviewers,
because
with
you
know,
with
the
changes
of
cilium
and
the
system,
I
really
want
to
support
at
least
join,
and
then
maybe
we
can
start
thinking
about
how
to
do
this
for
upgrade,
because
we
have
the
same
situation.
There.
A
D
A
A
D
A
D
B
So
this
is
like
one
of
the
side
effects
from
the
effort
of
trying
to
like
keep
parity
between
all
component
configs.
So
basically,
it
makes
sense
for
qubit
qubits
conflict
to
be
fetched
during
join,
but
it
doesn't
make
sense
for
queue
proxies.
But
since
there's
these
are
both
treated
by
the
same
interface,
which
is
the
component
config
interface
of
cube,
ATM
they're,
simply
just
fetched
up
when
whenever
we
actually
fetch
configuration
from
the
cluster
and
if
some
of
them
fails
it.
Basically
we
just
bail
out
and
say
we
couldn't
fetch
something
I.
D
B
A
And
also
because
a
question
whether
we
should
even
pour
the
couplet
confirmation
from
the
cluster,
because
it
makes
sense
for
mod
replicas,
but
maybe
not
for
nodes
that
have
a
completely
different
set
up.
I
have
a
major
disagreement
with
some
some
of
the
principles
of
the
corporate
confirmation
general.
But
this
time
I
have.
D
A
A
A
A
A
Yeah
so
I
was
here:
I
object,
yep
this
one
I
I
have
a
look
at
that
these
multiple
time
secure
of
had
any
problems,
but
I
requested
that
the
contributor
modifies
slightly
the
sorry.
Let
me
finish:
I'm
requested
that
he
should
they
should
modify
the
naming
of
the
function
to
basically
target
company
config
kinds.
A
A
A
A
A
A
Youtube's
matrix
to
the
component
base
upholstery
I
found
out
that
this
component
base
is
serving
us
the
purposes
also
as
an
important
repository
from
demons
in
general
or
server
utilities.
It's
not
only
about
genetic
components,
configuration
or
or
ukyo's
for
competency
or
comprehensive.
It's
also
about
server-side
stuff,
and
apparently
this
was
in
the
cable
therapy
session
here
with
launch
cocaine
and
also
I
think
mostly
spoke
with
stubborn
bodies,
but
yeah
I
guess
compress
the
cube
ATM
that
imports
the
component
base
gold
imports
are
going
to
allow
them
to
skip.
A
A
C
You
have,
please
tell
me,
but
yeah,
oh.