►
From YouTube: kubeadm office hours 2020-02-19
A
A
B
C
B
A
A
/
cover
illustration,
mapping
and
Fabrizio
created
a
housekeeping
task
here
for
this
when
help-wanted
track
an
issue,
but
we
didn't
get
any
contribution
for
that.
So
I'm
just
going
to
set
the
PM's
to
me
to
fix
the
problem,
it's
a
basically
a
CD
when
it
when
it
upgrades
it's
looking
for
a
mapping
between
119
and
a
particular
HD
version,
but
though
this
mapping
does
not
exist
yet
because
the
119
tag
is
from
yesterday
pretty
much
so
I'm
going
to
fix
that
to
be
alright.
I
wanted
to
have
4
bits
of
the
code
to
discuss.
A
D
D
A
A
It's
unresolved,
the
cap
can
still
be
merged
in
provision
of
state,
but
there
has
to
be
this
unresolved
tag
and
between
the
original
tags.
You
can
include
information
from
previous
discussions,
while
the
PR
was
still
in
progress
and
I'm,
not
sure
exactly
sure.
How
are
you
supposed
to
dump
the
discussion
in
this
HTML
format,
but
this
is
one
way
of
handling
the
errors
of
discussions
in
caps.
I
think
there's
more
to
this
I
think
the
date
format
is
changing
if
I
haven't
checked
exactly
how.
A
The
whole
template
is
changing,
so
maybe
next
time
we
create
a
cap,
we're
going
to
see
what
has
to
be
done.
It's
not
clear
to
me
if
existing
caps
have
to
be
changed.
I
think
the
data
format
has
to
change
at
least
for
the
old
caps
I
haven't
looked
at
this,
but
I'm
just
adding
this
as
a
PSA.
If
somebody's
interested,
please
see
our
cookies.
E
A
Basically,
though,
this
was
created
because
I
think
folks,
like
Tim,
coking
and
John
day,
America
were
looking
at
a
lot
of
caps
and
they
found
out
that
it's
really
hard
to
track
previous
discussions
and
problems
like
that.
So
they
decided
to
have
this
change
in
the
cap
process
and
as
our
people
agreed.
C
A
I'm
going
a
lot
of
people
are
going,
it's
not
clear
if
there
will
be
problems
with
the
virus
yet,
but
so
last
time
in
Barcelona
our
face-to-face
was
sitting
on
a
table
on
a
pretty
big
table
and
discussing
stuff.
I.
Think
this
face
face
was
productive,
I,
don't
think
we
have
to
book
a
room
or
something
like
that.
Yeah.
C
E
C
C
C
A
A
C
We
would
just
need
to
persist
the
declared
configuration
that's
considered
current
from
coubertin's
perspective
into
the
cluster,
and
then
that
allows
us
to
do
declarative
state
changes
so
like.
If
you
wanted
to
change
the
version
or
remove
an
add-on,
then
the
that
would
be
the
easiest
way
to
do.
That.
C
There's
no
like
grouping
construct,
that's
formal
for,
like
related
objects,
inside
of
kubernetes,
so
like,
and
no
way
to
like,
attach
metadata
to
them.
Things
like
the
application,
custom
resource
and
some
of
the
things
we've
proposed
in
add-ons
solve
that
problem,
but
are
not
generally
available
and
yeah
okay.
So
the
basically,
if
we
kind
of,
can
trust
how
it's
bundled
from
outside
of
the
cluster
and
then
reference
it
in
a
reliable
way.
And
then
we
store
those
references
in
a
config
nap.
C
A
I
see
with
other
mokopuna
config,
they
realized
that
we
are
facing
the
problem.
If
the
other
one
is
stellar
itself
suddenly
does
not
support
a
particular
version
of
a
GDK
of
its
own
G
became
how
we
faced
the
same
problem,
the.
How
are
we
going
to
upgrade
the
users
without
including
the
internal
types
of
that
on
having
said
Canadian.
C
And
then,
regarding,
like
users,
upgrading
their
own
configurations,
this
is
a
general
config
problem
that
we
have
right
now
since
tools
to
help
users
convert
things
don't
exist.
We
have
the
defaulting
logic
which
prints
like
pretty
prints.
The
defaults,
but
I
am
very
invested
in
helping
solve
the
general
component
config
problem
for
assisting
users
and
upgrading
their
configs.
D
C
Ultimately,
if
you
want
a
user
to
be
able
to
reference
an
OCI
image
or
a
git
repo,
which
is
like
something
that
like
is
fetched
over
the
network,
then
you
need
something
that
points
to
it
and
then
even
like
a
directory
on
the
file
system
like
you
could
have
a
canonical
directory
where
every
subdirectory
is
for
curst
and
that's
just
the
strictest
simplest
form
of
Unix.
He
config.
Yes,.
A
C
A
C
Yeah,
I'm,
I
am
very
positive
about
helm.
There's
really.
It
really
has
a
select
number
of
weaknesses,
but
I
think
it's
an
incredibly
strong
tool
that
has
unparalleled
user
adoption
in
the
kubernetes
space
and
I'm
I'm,
for
you
know
using
helm
in
places.
I
think
customized
has
received
a
lot
of
sponsorship
directly
from
the
project
and
is
a
simpler
and
more
restrictive
tool.
It's
not
as
powerful
if
you
want
it
to
say,
like
update
the
image
tag
of
something.
C
C
C
C
It
is
the
direction
that
I
think
a
lot
of
us
in
the
group
would
like
to
go
is
to
actually
have
a
standard
way
to
bundle
manifests
with
a
way
to
get
to
OCI
images
from
docker
registry
compatible
images
and
yeah
I
mean
calling
those
like
onto
the
box.
As
you
know,
whatever's
inside
of
the
add-on
configuration
inflating
those
things
and
pulling
them
would
be
good.
C
A
A
A
See
bundles
somebody
sent
a
PR
foresee
a
bundle
sporting
comedian,
but
this
is
great.
However,
the
controller
management
was
not
superseded
by
those,
so
it's
not
going
to
work
and
apparently
they
tested
it
in
a
way
that
bypassed
the
problem
so
I,
yes,
so
I
tried
to
explain
to
them
and
I,
don't
think
they
had
a
fauxhawk
I
think.
The
idea
here
was
that
they
should
basically
create
a
ticket
for
the
key
KCM
folks
to
investigate
potential.
You
see,
go.
A
A
A
A
B
G
A
Okay,
yeah
I
had
some
conversation,
so
I
guess
publisher
can
have
the
proof
lgt
about
this.
A
B
A
F
A
G
A
A
A
The
condition
by
seagal
for
ebay
DM
talks,
so
the
user
is
trying
like
different
things,
to
rotates
the
CA.
They
found
a
little
bit
of
a
problem
in
our
alpha
command
for
renew
certificates,
so
I,
better
change
here,
I'm
going
to
ping
Felicia
later
about
it.
It's
make
sure
that
our
ship
config
CA,
is
always
always
at
in
it.
G
I
think
that
already
commented
exactly
not
so.
The
idea
here
should
probably
be
to
just
replace
those
flags
with
Cupid's
component
completely
fields
when
such
fields
actually
arrived
and
not
just
have
a
single
yeah.
That's
going
to
go
for
it
and
read
everything
so
I
think
that
even
we
followed
the
work
completed
here,
we
won't
end
up
having
such
a
P
and
it's
simply
moving
stuff.
G
A
Which
was
the
third
one,
the
I
think
it
was
hot
sandbox
eh,
something
like
this,
oh
yeah.
So
this
is
dr.
shim,
it's
what
you
studying
here,
it's
not
in
the
config,
so
only
the
pot
sandbox
is
actually
what
presents
I
think
everything
else
can
be
so
somebody
somebody
started
working
on
a
beer
for
that,
but
they
closed
it
in
it
was
almost
complete.
A
G
A
G
Those
API
groups
I
think
that
right
now,
I'm
thinking
on
more
broadly
interface,
that's
probably
going
to
result
in
several
small
hopeful
commands,
which
is
basically
allowed
for
some
better
UX
around
this.
So
it's
not
just
this
table.
It's
probably
going
to
end
up
in
an
output
man
to
upload
some
convicts
and
want
to
download
the
offending
convicts.
A
It's
fine.
If
we
I
imagine
a
scenario
where
somebody
might
you
know,
maybe
it's
a
typo
in
there
llamo,
but
at
some
point
someone
might
have
imagined
different
couplet
kinds
and
one
of
them
is
versioned
with
v1
alpha.
Sorry,
if
you
want
beta
2
and
then
what
other
one
is
version
with
v1
be
the
tree.
So
it's
a
it's
a
complication.
Well
cube.
Adium
has
to
basically
track
different
versions
for
different
kinds
in
this
case.
G
G
G
What
we
like,
what
we
should
do
here
is
evaluate
the
UX,
get
some
initial
concepts
in
place
and
probably
enrich
that
message
came
with
some
additional
commands,
for
example,
with
some
advice
on
what
command
to
run.
If
API
group
needs
a
user
action,
what
command
to
run
to
fetch,
unsupported
configuration
or
s,
clip
deleted
or
and
and
what
command
to
if
a
user
has
a
newer
version
of
the
configuration
in
Chinese
to
uphold.
A
G
A
G
Here
SK,
the
idea
is
to
either
around
for
the
user
to
delete
their
own
users.
Bite
and
I
mean
users
fight,
because
only
users
fight
configurations
should
require
a
manual
user
grade
or
deletion,
so
the
users
can
actually
opt
in
to
either
supplied
or
modified
configs
and
allow
qadian
to
regenerate
them
with
the
new
ones
or
should
actually
be
able
to
see
what
they
are.
All
configurations
look
like
and
then
just
go
in
and
supply
newer
versions
of
dogs
which
are
supported
by
QA.
A
G
A
There's
also
the
option
I
guess,
but
it's
it
requires
some
boilerplate
too.
If
the
user
Yama
on
disk
is
passed
this
command
and
if
the
command
is
aware
of
the
new
kind
and
all
the
fields,
we
can
do
a
field
thief
of
sorts,
which
is
not
great,
but
we
can
say
hey.
We
found
that
the
new
corporate
coffee
has
these
fields,
but
I
don't
like
this
in
that
much
good
yeah.
G
G
A
So
this
is
gachi
who
is
working
with
memos
from
people
to
join
VMware
recently
he's
making
a
change
in
our
flags,
so
our
flags
for
the
couplet
he
wants
to
make
them
always
dependent
and
I.
This
is
fight
against
this
again,
a
discussion
with
rusty.
Do
you
think
that
we
should
use
the
underscore
OS
pattern,
or
should
we
use
runtime
checks.
G
A
A
So
the
Sierra
setup
is
something
that
Windows
needs.
It's
unique
because
of
the
powers
image
for
dr.
shim,
then
the
Linux,
what
the
Linux
logic
nice,
also
this
year,
I
set
up,
but
also
the
resolve
coffee,
which
one
day
we
want
to
move
to
the
copilot
configure
the
cool
by
separating
the
docker
ship
stuff.
Do
you
mean
like
a
separate
function,
maybe
with
a
runtime
check,
yeah.
G
A
separate
function
that's
going
to
produce
basically
the
same
as
this
one
produces
and
in
the
end
we
just
merge.
Those
lumps
idea
it
doubles
should
not
be
overlapping.
Then
I'm
not
entirely
sure.
Why
there's
a
lot
conf,
it's
actually
inside
of
the
criminal
configuration
as
of
now,
but
it's
mostly
dr.
shim,
but
probably
the
capabilities
using
it
in
some
other
way
too.
G
But
I
think
that
if
those
facts
are
supported
as
their
own
Linux
and
they're,
basically,
if
they
are
supported
in
the
same
way
on
Windows
2,
we
can
simply
leave
those
bugs
there
and
just
supply
them
with
the
correct
those
posters.
For
example,
this
here
I
set
up
curse
planes
here,
I
socket
unlimited,
supplying
a
UNIX
domain
socket
on
Windows,
yours,
plane,
I
think
a
named
pipe.
A
A
G
Yeah,
so,
okay,
let
let's
go
in
this
direction
and
see
how
it
ends.
As
I
said,
this
is
supposed
if
some
of
these
facts
are
really
doctor,
she
in
air
entangled
in
there.
So
what
we
mean
the
some
day
window
crucial
separates
from
the
key
blood
is
basically
the
those
functions
going
to
end
up,
probably
the
same
or
just.
A
It's
also
not
clear
to
me
what
what
is
the
state
of
the
general
work
of
creating
a
multi
OS
image
for
the
pause
container.
I
guess
this
worked.
Work
was
closed.
I
think
this
should
erase
licenses
of
the
Microsoft
images,
but
the
base
image
is
on
nano.
Nano
server,
so
I
think
there
was
a
problem
with
that.
I,
don't
think
he
take
a
map
associates
to
work
there
or
something
like
that.
I
don't
know
the
details,
but
from
now
I
guess
we
have
to
use
separate
constants.
A
D
A
G
And
I
think
that
basically,
we
shouldn't
be
dealing
with
the
cube
ADM.
This
is
a
CMS
stuff
for
the
doctrine
case
we
just
have
to
because
it
is
basically
being
dealt
with
by
blood.
So
I
think
that
probably
this
list
is
somehow
acceptable,
at
least
on
my
side.
I,
don't
see
any
more
clear
solution
to
this,
especially
if
it's
false
image
from
Windows
is
not
about
it.
Jesse
a
poster,
probably
the
best
way
to
deal
with
this
just
walk
us
in
GC
our
repository.