►
From YouTube: kubeadm office hours 2019-10-23
A
A
A
Quicker
cut
off
per
team
and
for
the
other
that
that
means
that
the
last
week
meeting
we
are
trying
to
manage
properly
the
unique
sequence
that
particular
to
make
corporate
configurations.
Sigma
has
the
configuration
of
the
other
nodes
and
basically
boil
apps
to
to
fix
up
the
kuba
data
doc.com
file
in
order
to
use
the
same
certificate
that
kubera
twitter
is
going
to
rotate,
and
we
we
made
a
couple
of
PR
triangle,
an
approach
sure,
but
it
didn't
work
either.
A
So
last
week
we
decided
to
add
a
new
face
and
at
the
end
of
the
quad
main
sequence,
and
this
face
is
going
to
basically
fix
up
the
Kampai
in
order
to
make
the
rotation
the
certificate
rotation
possible.
Also
on
the
first
contour
pain,
node,
and
we
are
creating
a
new
phase
with
one
phase
for
google
rotation
and
also
there
will
be
another
phase
for.
A
A
Another
ratio
which
is
pending
and
that
that
is
a
white
listing
bottle
a
was
the
the
pending
question
here
is
how
to
code
this.
With
his
face
no
bomb
a
proposal
was
cuba,
litter,
ad
and
rotate.
We
photography
is
called
the
rotated
clients.
Earth
I,
don't
have
a
strong
opinion.
I
made
a
counterproposal
saying
that
we
should
be
more
consistent
to
the
other.
A
D
D
E
A
A
Kubernetes
node
with
an
external
entity-
and
this
is
a
bug
and
we
agreed
that
I
agreed
with
dopamine
that
this
should
be
fixed.
The
second
case
was
check,
aspiration
is,
failing
is
basically
they
denote
where
the
user
is
tallied
tcd.
So
if
they
are
not
kubernetes
node,
but
they
are,
it
is
in
node,
created
with
KU
between
phases.
C
A
C
I
wanted
to
have
a
more
in-depth
discussion
with
this
group
about
importing
of
Covidien
types
and
the
issues
that
people
had
run
into
with
coaster
api
and
potentially
try
to
solve
like
enumerate
those
issues
and
potentially
try
to
solve
some
of
them
for
B
1,
beta
3
I'm
generally
concerned
when
downstream.
If
people
cannot
easily
vendor
the
the
types
that
you
would
expect
to
be,
composable
right,
I
do
know
that
there
are
some
transitive
depth
issues
that
exist
and
we've
been
slowly
kind
of
piecing
or
tearing
apart
tearing
them
apart.
F
You
want
me
to
chime
in
Tim
sure
so.
I
I
know
that
when
we
were
working
on
Cluster
API
alpha-2
pulling
in
the
cube
ATM
types,
we
ended
up
having
to
copy
them
into
our
code
because
we
needed
to
be
able
to
leave
certain
fields
empty
or
omit
them,
which
I
believe
was
acceptable
because
cube
EDM
I
think
was
defaulting
them.
But
we
weren't
able
to
get
that
I.
Think
if
I
remember
correctly,
the
types
themselves
required
values
to
be
filled
in
and
we
couldn't
even
just
we
couldn't
submit
our.
F
We
couldn't
submit
our
object
to
the
API
server
and
have
it
passed
validation
because
the
stuff
wasn't
getting
defaulted.
The
way
that
we
needed
it
to,
and
so
we
ended
up
copying
the
types
and
adding
some
of
it
empties,
where
it
helped
us
out
and
I
believe
that
Chuck
made
some
little
requests
or
opened
some
issues
about
that.
I
think
that
was
how
it
went
so
some
combination
of
too
strict,
too
strict
validation
or
not
getting
the
defaulting
that
we
needed
or
both
I
can
get
the
full
details.
F
Do
I
have
a
second
field
which
is
V
1
beta
2
in
it
that's
got
the
V
1
beta
2
package
do
I,
do
it
as
a
unstructured
data
and
do
my
best
to
allow
the
user
to
put
in
wherever
they
want,
which
basically
means
I
can't
use
open,
API
validation,
it's
just
kind
of
a
mess
when
there's
multiple
pre-release
versions.
I
would
imagine
that
once
it
gets
to
be
one
that
V
one
would
probably
be
supported
for
a
long
time,
and
people
could
rely
on
that
and
not
have
to
worry
about.
C
G
C
F
F
F
There
is
some
issue
with
CRD
structural
schemas,
open
api
validation.
The
stuff
that
cube
builder
generates
that
if
you,
if
you
try
to
have
a
widget
that
has
its
its
normal
object
meta
at
the
top,
but
then
you
have
somewhere
in
your
object
graph
or
in
your
fields.
Another
object
meta.
If
cube
builder
doesn't
like
it.
C
So
I
think
we
have
solutions
for
so
far.
We
have
enumerated
three
things,
then,
actually
so
in
the
doc.
The
first
one
is
being
worked
on
would
be
one
beta,
3,
the
third
one,
the
object
made
every
one
is
being
addressed,
the
second
one,
the
multiple
API
versions
and
embedding
types
with
inside
of
the
larger
others.
Other
types
I
don't
exactly
know
how
we
want
to
handle
that
I'd
have
to
think
how
kubernetes
does
it
today,
because
we
do
do
that
all
over
communities
and
whether
or
not
it
is
only
a
core
v1
types.
Well,.
F
Yeah,
so
the
fact
that
it's
most
of
those
are
at
v1,
you
don't
have
this
flip-flop
or
multi
version
issue
and
I
mean
anyplace
else
that
is
using
a
deployment
or
something
from
cube.
80Mm
is
going
to
use
a
specific
API
version
and
not
multiple,
and
if
needed
it,
you
you
could
there
would
be
conversions
either
done
automatically
or
you,
as
a
user,
would
be
instructed.
Well,
if
you've
got
an
older
thing,
you
need
to
convert
it
to
the
new
one
before
we'll
take
it.
So.
C
If
I
were
super
old
school
and
I,
just
ignore
API
machinery
for
where
what
it
is
today
like
just
just
ignore
just
say
it
doesn't
exist.
I
would
create
a
discriminated
union
of
a
type
which
would
have
both
types
embedded
inside
of
it
and
I
would
call
it
it's
like
an
envelope
letter
idiom
could
be
called
cube,
ATM
types
for
Covidien
config,
which
would
contain
the
multiple
versions
of
b1
beta
1
and
B
1
beta
2.
They
would
use
that
side
of
my
API.
C
C
F
C
F
F
Which
is,
if
you,
it
is
different,
I
mean
it's
the
same,
like
it's
the
same
name
in
that
result,
you
can
do
it,
but
when
you
have
a
union
of
multiple
types
where
only
one
of
them
is
present,
those
types
generally
are
different
with
what
you're
proposing
it's.
It's
always
gonna
be
a
cube,
idiom
config,
but
it
might
be.
C
We
should
expose
that
generic
type
so
that
people
that
are
consuming
it
don't
need
to
care
so
like
as
part
of
our
types
file
that
we
generate.
We
have
the
specific
versions
there
and
we
have
the
generic
wrapper
that
allows
people
to
consume.
It
I
think
the
onus
is
not
necessarily
on
you
but
might
be
on
us
right
so
might
be
Kure
DM.
It's
like
the
distinctions
between
us
and
whatever
the
onus
might
be
on
Canadian
types
to
have
that
wrapper
object
for
general
consumption
because
it
might
be
generally
useful,
but.
C
F
You
saying
that
this
wrapper
type
would
have
helper
functions
to
get
all
the
individual
thing
getting
set.
Well,
maybe
not
set,
but
at
least
get
so
from
a
consumption
standpoint
once
everything
once
you've
filled
in
that
it's
a
v1
beta,
3,
init
config,
if
I
say,
get
some
field
that
would
go
to
the
right
place
and
I
don't
have
to
have
I,
don't
have
to
write
in
my
code
check
this
version,
then
that
version
and
some
other
version.
That
would
be
the
ideal
scenario.
That's.
H
H
F
Well,
so
the
internal
type
is
never
a
versions
type,
it's
it's
a
point-in-time
version,
but
doesn't
have
an
API
version
associated
with
it.
So
it's
the
it's
the
hub
through
which
you
would
convert
to
beta
1,
beta
2,
beta,
3,
etc.
But
the
reason
that
it's
not
exported
is
because
it
doesn't
have
a
stability
guarantee
behind
it.
Yes,.
F
H
D
C
So
I
propose
the
following
because,
like
what
I
want
to
do
is
get
a
list
of
key
detractors.
Instead
of
us
trying
to
bike
shed,
I
can
open
an
issue,
and
hopefully
Andy
can
give
me
the
concrete
example
that
which
was
the
configuration
through
control,
plane,
I,
think
or
it
was
part
of
the
construction
bag.
I
have
a.
A
Comment
here
that
probably
a
good
approach
to
this
is
dead
from
the
cluster
of
the
I
perspective.
In
not
all
the
detail
of
the
cubed
mini
pii
are
rare,
a
lot,
but
only
few
of
them,
so,
for
instance,
it
is
a
crass
therapy
is
interested
in
understanding
if
there
is
a
token
or
if
the
joint
configuration
is
for
is
for
joining
a
cluster
node
or
a
worker
node.
G
G
F
Well,
we
do
need
to
be
able
to
specify
things
like
the
flags
or
component
config
when
it
comes
around
for
the
controller
manager
in
the
cube
lid
and
the
API
server
and
whatnot.
So
it's
I,
don't
think
trying
to
restrict
what
people
are
able
to
set
in
the
anit
cluster
and
join
configs
right
now
is
something
we
should
really
look
to
do,
but
if
I'd
be
interested
to
learn
where
component
config
is
going
and
if
that
would
make
things
easier,
yeah.
A
F
C
Yeah
I
think
I
think.
Instead
of
continuing
on
bike
shed
approaches,
we
can
open
an
issue
and
try
and
get
resolution
on
I.
Consider
this
to
be
pretty
high
priority.
I
do
not
want
to
see
weird
type
proliferation
and
separate
conversion
utilities
all
over
the
place.
I,
don't
think
that's
the
best
interest
of
plus
JP
I,
don't
think
it's
the
best
interest
of
qubit
again
I
think
it's
better
to
try
and
solve
this
problem,
which
is
generally
hard
and
try
to
figure
out
a
path
forward.
C
I
A
Finding
the
document
DDA's
to
hide
all
the
detail,
were
it
possible
to
to
keep
things
at
a
level,
so
the
first
phase
will
be
per
flight
and
yet
before
the
phase
there
will
be
the
profit.
But
this
is
out
of
the
phases
within
a
mistake.
Then
we
have
proof
right
as
usual,
then
we
have
a
unique
phase
that
manager
they
are
great
for
all
the
components
for
the
contracting
companies
entity
CD
and
we
are
keeping
this
as
unique
phase.
A
Then
we
are
grouping
all
the
posted
grade
faces
or
all
the
poster
grade,
steps
like
epidural
configuration
and
write
down
the
new
cover
that
config
file
and
reapply
buck
rules.
We
are
grouping
everything
in
a
you
know
nod
in
another
phase
without
any
surfaces
and
finally,
we
are
keeping
advanced
as
a
separate
phases
with
two
surfaces,
so
the
user
can
skip
to
the
proxy
or
accordion
s.
D
A
E
A
So
here
we
have
a
unique
function
which
is
post
upgrade
that
does
a
bunch
of
things
upload
a
configuration
he
create
the
call
the
config
map
for
the
new
google
component
configure,
so
they
call
kuba.
Let
me
know
severe
shaman
config
map,
it
rised
the
Cooper
configuration
and
upgrades
the
local,
the
local
Kubla,
to
use
the
new
version
of
the
configuration
unnoticed,
the
denotes
to
eat
rivera
upgrade
after
days
the
notation
of
the
node.
A
I
C
J
A
J
I
I'm
not
sure
if
we
actually
want
to
add
so
at
first
I
really
liked
the
idea,
because
it
just
provides
a
better
user
interface.
You
know
if
it
doesn't
just
fail
because
of
a
hiccup
in
the
network,
but
then
I'm
not
sure
if
we
really
want
to
add
all
the
complexity
to
this,
so
I
want
your
guys
thoughts,
my
kind
of
in
now
a
wash
on
whether
we
do
it
or
not.
C
Even
inside
of
the
client
code
for
anything
that
talks
to
the
kubernetes
cluster
has
a
set
of
exponential
decay
back
off
stand
eventually
time
up.
So
those
having
that
behavior
be
consistent
seems
rational.
I
do
think
that
it
would.
You
can
hide
it
behind.
There's
utility
functions
that
do
this
forever.
What
they're
called
they're
in
the
utils,
but
you
can?
You
can
have
that
capability.
A
C
C
J
D
Didn't
we
discuss
something
like
that?
I
remember,
there
was
a
TR
was
opened
a
while
ago
that
checks
against
America.
Is
it
related
to
this
tricks?
What
sorry?
It
was
a
PR
I
I,
don't
find
it,
but
it
was
a
PR
that
was
opened
and
reviewed
by
Ross
on
me
that
checks
against
HTTP
toad
I,
don't
recall
if
it's
still
open
I
have.
A
A
This
is
a
good
recommendation
for
the
user,
maybe
we
varieties
in
the
website
in
case
there
are
potential
problem,
a
network
problem.
It
is
better
to
procure
image
before
running
convenient.
There
is
also
a
command
to
help
refer
base
and
in
case
of
cloud
provider,
images
can
be
pre-loaded
in
the
template
of
images.
So
there
are
many
way
to
do
this,
and
so,
but
but
nevertheless,
I
don't
see
probably
nothing.
They
dare
try.
A
A
J
Would
open
the
117
miles
from
because
what
I'm
doing
nowadays,
you
just
put
everything
in
the
mouth
on
which
means
that
it
needs
triage,
and
if
you
think
that
it's
not
appropriate
117,
it's
now,
it's
the
time
to
start
moving
things
to
the
next
milestone.
The
most
118
already
exists.
Also,
we
have
next
applying.
A
J
A
Okay,
this
one
about
the
the
Czech
aspiration.
I
have
to
comment.
I,
don't
know
if
you
were.
If
you
join
when,
when
it's
custard
the
year
is
to
fix
the
defect,
for
Cooper
mean
the
Farakka
would
mean
for
checking
aspirational,
mean
node
with
us
tonight
ECD,
but
not
to
not
support
the
check
expiration.
On
that
easily
note.
J
A
E
J
C
E
C
J
This
is
pramana
D,
yes,
this
is
also
accelerate.
City
I
can
summarize
this.
So
basically
the
user
is
doing
config
images
full,
but
they're
not
passing
the
config,
so
the
request
they
have
is
to
fetch
the
cube
ATM
coastal
configuration
from
the
config
map
and
determine
if
the
cluster
is
running
external
HDD,
and
but
my
objection
to
that
is
that
this
command
can
be
used
without
the
cluster
Abell
running.
D
J
A
J
Yes,
so
it's
already
possible
to
pass
config
from
which
already
the
pool
command
can
determine
if
external
HDD
is
used
or
not.
But
do
you
want
to
pass
only
the
version
and
basically
the
user
is
proposing.
Hey
I
want
to
fetch
the
config
map
and
our
a
proposal
to
fix
the
book
where
whether
we
should
check
expiration
for
EDD
it's
similar,
because
we
you
propose
there
to
fetch
the
config
map
of
the
end.
So
this
is
the
same
the
same
stuff,
so
we
have
to
be
consistent.
J
A
Okay,
let's
table
the
discussion
and-
and
we
try
to
think
of
line
ie
to
be
honest,
I
was
thinking
about
the
one
about
check
expiration
because
also
check
his
parish.
Erisa
is
a
comma
and
the
debt
should
be
capable
to
run
without
a
working
cluster,
because
it
is
your
last
ratio
in
case
of
the
class
that
is
passed
at
the
first
certificate.
So
basically
it
is
a
chicken
era.
If
I
ask
the
user.
If
I
ask
the
comment
to
us
sister,
a
cluster
I
need
a
certificate.
The
valid
head,
I
cannot
renew
them.
A
A
D
F
J
A
E
J
Basically,
what
what's
Daniel
wants
here
is,
if
you
enable
like
it's,
not
official,
it's
not
official
decision,
we
don't
have
one
it,
but
I
think
what
we
should
do
here
is
it
for
have
velocity
levels.
We
should
dump
the
the
caustic
inflation
each
configuration,
whatever
structures
we
just
read
from
by
the
user
from
the
user,
and
we
default
it.
So
simply:
dump
these
configurations
to
standard
error
using
K,
walk
or
something
yeah.