►
From YouTube: kubeadm office hours 2019-06-26
A
A
A
A
A
ruffler
has
been
helping
us
with
a
lot
of
core
elements
of
qadian
on
a
regular
basis,
and
maybe
we
can
vote
on
him
to
be
an
approver
I
said
the
QE
has
been
very
active
in
communities
in
Judea.
He
made
it
to
the
top
100
list
of
contributors
to
KK
and
he
also
has
been
helping
the
Canadian.
But
writer
is
writer,
so
I'm
plus
one
on
all
these,
but
we
only
have
Tim
as
the
polar
I
guess.
We
don't
have
quorum
to
vote.
B
B
A
Okay,
so
next
on
the
agenda
is
basically
we
wanted
to
discuss
the
cube
alien
dependency
tree,
but
we
don't
have
all
of
the
people
who
want
to
help
with
this
and
I
mean
rusty
is
on
vacation.
I,
don't
see
Eddie's
here,
okay,
but
I
also
wanted
to
have
dims
here,
because
I
have
a
bunch
of
questions
for
staging
in.
D
A
A
B
I
would
I
would
prefer
to
skip
staging
entirely.
There's
no
there's
no
benefit.
If
we,
if
we
can
isolate
the
dependencies
and
make
dims
and
short
and
happy
so
that
we
have
reduced
dependency
tree,
that's
a
general
problem.
We
should
solve
anyways,
then,
once
we
get
to
the
point
where
we're
satisfied
that
we
can
move
on
a
tree
just
wholesale,
we
can
just
pull
the
trigger
and
just
do
it.
I
think.
B
D
A
E
C
A
Aging,
but
something
might
going
to
the
next
question,
I
guess
is:
can
we
leverage
the
publishing
pots
in
a
way
that
if
we
move
to
k
/q
baby
M,
can
we
use
the
publishing
pot
for
tags
and
branches
only
to
automate
the
process
of
that?
If
we
decide
to,
you
know,
continue
matching
the
Canadian
version
to
the
community's
version.
D
Yeah
I,
don't
think
yeah
I,
don't
think
it
supports
that
either
I'm,
not
even
sure
publisher,
but
does
like
the
actual
release.
Tagging
like
I
know
it
sinks
branches
as
well,
but
I'm,
not
sure
I.
Think
the
the
tagging
but
I
know
Clank.
Oh,
it's
still
tagged
from
like
a
natural
human.
So
like
I'm,
not
sure
that's
that's
something
popped
in
shabbat
can
do,
but
that's
something
that
can
probably
you
could
probably
open
an
issue
and
see
if
they
can
support
that.
B
Once
once
the
I
say
at
the
dependencies
because
of
the
way
Covidien
was
written
to
grab
from
GCS
buckets
like
there's
no
hard
dependency
like
once,
we
isolate
the
actual
code
graft
dependencies,
there's
no
actually
hard
dependency
on
I'm,
KK
or
anything
there,
because
we
just
rely
on
it
on
a
rally
point
which
is
a
GCS
bucket.
So
if
we
were
totally
to
be
outside
a
tree,
there's
there's
a
lot
of
advantages,
at
least
from
a
release.
B
A
Yeah
something
something
about
the
kubernetes
release
process
of
with
regards
to
some
project
is
that
I
think
is
very
healthy.
To
have
a
you
know,
a
solid
release
cycle
and
you
know,
keep
yourself
under
the
the
pressure
to
release
every
three
months
and
if
we,
even
if
we
don't
elect,
know
exactly
much
attack
in
KK
I
think
we
should
still
continue
to
match
the
community's
release.
This
also
give
us
gives
us
a
very
simple
support
matrix,
because
currently
we
support
this
version
of
core.
A
B
I
agree,
I
think,
there's
there's
a
huge
benefit
towards
having
the
cadence
match.
It
doesn't
need
to
be
entirely
lockstep.
We
could
have
some
slack
there,
because
I
almost
guarantee
you
that,
like
every
release
of
Kay
Kay,
always
comes
with
like
oops
like
really
fast
afterwards
and
not
being
subject
to
those
whoopsies.
Actually
you
know
let
give
it
a
week.
Timeout
or
something
like
you
know,
add
one
week
to
the
end
of
the
release
and
then
you
get
the
artifacts
is
not
a
bad
thing.
Unless
somebody
wants
to
try
it
out
and.
E
B
A
Yeah
apart,
services
are
going
to
be
tricky
because
we
do
have
the
freedom
to
pull
any
image.
We
want
for
the
control
plane,
but
we
have
to
figure
out
the
passionless
cadence
of
qadian.
Is
it
going
to
be
country
or
completely
arbitrary?
Or
if
you
want
to
still
match
the
kubernetes
release,
cadence
of
patches,
yeah.
B
Well,
I
think
apartment
I
think
a
precondition
to
to
finally
pulling
the
trigger
is
like
we
need
to
get.
We
need
to
make
sure
that
some
of
the
other
apparatus
we're
talking
way
far
down
the
road,
in
my
opinion,
like
118
Plus,
because
if
we
get
to
the
point
where
we
need
to
make
sure
that
the
the
dependencies
that
we
have
don't
get
shifted
up,
underneath
our
without
our
purview
be
able
to
watch
it
right,
and
currently
we
have
that
problem.
Where
changes
get
made
subtly
to
other
locations
that
would
drastically
affect
us.
B
We
want
to
make
sure
that
that
it's
not
possible
in
the
new
world
order,
so
I
think
I.
Think
we'll
worry
about
that
when
we
get
there
I
think
in
the
near
term,
just
isolating
dependencies
is
a
reasonable
effort
for
us
to
try
and
shoot
for
and
maybe
along
the
way
we
can
produce
some
documentation
on
how
we
could
envision
ourselves
moving
forward
in
a
new
world
order.
So
that
way,
when
that,
when
we
do
hit
that
point,
we
have,
we
have
all
the
eyes
that
identities
crossed.
A
A
A
So,
ideally,
today,
I
wanted
to
discuss
some
of
the
more
problematic
from
you
know:
areas
we
have
so
maybe
we
should
start
with
the
public
types,
because
that's
the
most
requested
thing
that
people
want
and
basically
this
these
are
the
public
terms
correctly.
The
problem
with
the
public
types
is
that
they
import
the
internal
types
because
they
support
defaulting
and
commercial.
This
is
a
part
of
it.
The
machinery
that
it
it
requires
you
to
have
some
functions
and
machinery
without
the
public
type
package
itself
to
support
conversions.
A
I
can't
say
that
I
certainly
agree
with
that,
but
the
way
we
should
do
it
is
basically
the
way
approxi
doing
it.
Thank
you
props.
This
is
a
very
tricky
proxy
right
now
they
they
only
move
the
types
to
gopher,
that's
that's
all
they
do
and
they
skip
the
movement
of
commercials,
but
for
cupid
limit,
because
the
question
like
where
should
we
put
this
because
we
currently
don't
have
a
location
for
that?
Apart
from
the
you
know,
maybe
kubernetes
keep
a
memory
for.
A
So
that's
why
I
asked
the
first
question
was
for
the
publishing,
but
like
okay,
if
we
start
doing
it
now,
maybe
we
want
to
have
an
automated
process
for
it,
but
it's
good.
It
is
going
to
milk
all
the
existing
content.
Communities
begin
for
the
sake
of
moving
the
you
know,
the
the
public
types
there
and
I
really
don't
know
about
how
to
approach
this
and
Fabricio.
Basically,
we
had
a
discussion
on
Australasia
publishing,
say
like
stop.
B
You
know
if
they
were
functioned
as
CR
D
is
separately.
You
know,
could
builder
make
sure
makes
certain
that
the
separation
of
concerns
for
the
types
that
you've
created
doesn't
drag
in
other
things,
so
you
can
import
them
cleanly.
So
I
must
be
missing.
Something
I
thought
this
was
totally
possible
already
in.
It
seems
like
it's
just
a
issue
with
how
we've
done.
However,
you
use
the
tooling.
B
A
On
the
conversion
would
go
in,
you
know
in
the
D
1
B
2
we
import
the
internal
Kapadia
type
and
if
somebody
wants
to
you
know
import
this
whole
package
as
as
is
the
it
is
currently
they're
going
to
drag
the
internal
configuration
of
key
medium,
and
we
don't
want
that.
So
what
keep
rocks
you're
doing,
I
just
need
to
type
the
URL.
Sorry.
D
A
B
A
And
these
are
the
actually.
These
are
the
internal
types
of
your
boxing,
the
couplet
okay.
So
that's
the
problem
right
there
that
that's
that's
like
a
separate
problem.
We
I'm
going
to
talk
about
it
later.
Like
correctly,
we
should
think
about
how
we
can
continue
this
pattern
that
you
proxy
have
to
bow
fully
these
files,
like
the
question
is
where
maybe
I
was
thinking
like
we
can
move
into
a
to
a
folder
on
the
community,
static
medium
and
but
the
question
is
how
how
can
we
sing
on
a
regular
basis?
B
A
A
problem
you
need
to
have
machinery.
Basically,
if
you,
if
you
create
a
version
package,
you
have
to
have
the
converters
inside
the
partial
package
and
the
converters
depend
on
the
internal
types
which
is
I,
told
us,
and
what
this
is
done
like
that
we
are
not
doing
anything
wrong.
This
is
already
the
schema
another,
the
mechanic
that
is
using
copulating
to
proxy,
that's
how
it
works.
So
so,
if
you
want
to
export
something
public,
you
just
have
to
copy
the
file
somewhere.
A
D
A
D
Gotcha
and
even
if
we
did
do
the
cut
to
external,
ideally,
we
don't
want
to
render
back
in
KK
right
and
you're
importing
the
cubelet
configs
api
I
think
I
sell
it
correctly,
which
is
probably
something
that
needs
to
be
addressed.
So
yeah
seems
like
a
like
a
component
based
working
group
problem.
First,
at
least
that's
how
I'm
seeing
it
because
they're
already
trying
to
prune
dependencies
there.
A
Yeah
part
of
the
problem,
which
one
I
understand
the
the
long-term
plan
for
keep
oxygen
couplets
intro
types
is
that
we
wanted
a
hub
repository
somewhere,
where,
basically,
you
can
have
the
internal
types
for
all
these
components
and
everybody
can
import
them.
But
this
is
conflicting
with
the
philosophy
of
API
machinery,
where
nothing
except
the
component.
The
component
itself
should
import
internal
types
and
basically
only
public
that
should
be
imported
by
third-party
components.
B
This
is,
we
might
want
to
talk
with
epi
machinery,
folks
about
fixing
some
of
this,
because
it's
the
way
CR
these
work
is
you
have
a
conversion
that
which
is
a
totally
separate
whole
thing
right,
and
that
way
your
conversion
web
help
could
be
built
out
separately
and
if
we're
gonna
move
things
to
component
config.
This
is
a
component
good,
big
problem
for
how
they
won't
represent
ppm
machinery
and
I.
B
B
There
are
many
idioms
that
we
could
impart
that
people
don't
have
to
import
this
specific
element
and
have
the
capabilities
for
doing
it,
be
hidden,
underneath
some
registration
function
or
something
right
so
like
there,
there
are
many
crafty
things
you
can
do
to
do
code,
hiding
that
prevents
you
from
having
to
export
this
conversion
mechanism
via,
like
you
know,
stick
it
in
an
if
function
as
part
of
an
it
for
this
module
right
automatically
registers
itself.
None
of
that
stuff's
important.
You
don't
see
it
right,
it's
hidden
behind
the
scenes.
B
A
Absolutely
we
ideally
I
wanted
someone
from
API
personally
to
be
joined,
to
join
the
meeting
today,
because
people
like
David,
EADS
and
Jordan
express
it.
We
are
doing
things
wrong,
but
then
again
we
don't
know
what
is
supposed
to
be
the
correct
way
and
also
this
seems
to
be
something
disagreements
there
and
also
I
disagree
with
the
current
way.
Api
machinery
works
with
regard
to
politics,
but
yeah
when
Felicia
is
back.
We
are
going
to
talk
about
tips
for
this,
because
it's
complicated
yeah.
B
B
A
So
in
the
list,
I
have
a
couple
of
items
here:
they're
separate
problems,
technically
the
internal
types
for
proxy
and
the
other
one
is
the
Cabiria
public
types.
So
basically
I
already
said
p1
to
the
public
types
of
QA
DM.
But
this
one
should
we
add
it
as
P,
2
or
P
1
as
well
I
mean
the
Adept
they're
connected,
but
it's
still
separate
efforts.
I.
D
Would
put
depending
on
like
the
kubernetes
specific
cubic
to
proxy
internal
types,
had
a
lower
priority
than
like
some
of
the
external
dependencies
like
like
docker
or
whatever
else
that
would
conflict
with
other
six
like
I?
Don't
know
how
trivial
it
is,
but
if
we
were
to
decide
priorities,
it
seems
like
I'm
blocking
a
dependency
dependency
issue
with
like
cubelet
or
key
types
is
easier.
A
D
B
A
B
D
D
A
A
A
B
B
This
this
is
a
weird
one,
because
I
would
expect
this
type
of
behavior
to
be
exposed
via
component
config,
which
means
that
it
should
be
publicly
exportable.
We
can
copy
it
from
the
time
beam,
but
I.
Don't
definitely
put
it
to
do
on
this,
because
right
now,
every
other
component,
except
for
the
API
server,
has
computed
config
style
thing
broken
up
for
it.
Yeah.
A
Again,
this
is
the
basically
the
time
importing
the
internal
types
of
the
couplets
and
same
for
the
cue
proxy,
but
these
ones,
okay,
manifest
parts
of
the
cues
in
the
parent
for
this
should
be
resolved
at
this
point,
I
think
the
cue
proxy
intro
types
now
this
we
import
make
easy
proxy
IP
vs
for
the
sake
of
running.
You
know,
IP
less
pre-flight
checks,
and
maybe
we
should
stop
doing
that.
B
I'm
trying
to
think
of
where
every
live
I
would
expect
it
to
potentially
live
in.
A
component
could
possibly
live
in
a
set
of
utility
functions
which
could
be
part
of
the
component
config
for
the
proxy
itself,
because
consumers
who
want
to
use
the
proxy
properly
from
a
component
configuration
perspective,
would
would
have
those
utility
functions.
B
We
should
we
should
talk
with
the
component
fate
config
folks,
because
it
makes
it's
kind
of
like
you
know.
Bootstrap
turkan
says
that
some
utility
functions.
You
should
have
that
stuff
co-located
with
anything
that
that
is
going
to
use
those
api's
right,
because
otherwise
it's
not
really
useful.
It's
kind
of
like
Quantico
client
was
it
is
the
exposed
types
along
with
all
these
utilities.
C
A
E
A
D
A
D
A
Need
to
I
need
to
look
at
the
package
again
because
I
already
looked
at
these
functions.
I
didn't
have
the
time,
so
maybe
maybe
it
fits
it
fits
through
to
be
moved.
Cuties
I,
don't
know,
okay,
this
one.
It's
already
resolved,
you
see
insane,
but
we
are
to
move
the.
Basically,
we
have
an
abstraction
layer
for
different
in
it
systems.
You
know
like
system,
D
or
the
windows
services.
So
this
now,
let's
enter,
keep
again.
I
think
this
was
a
fine
move.
A
A
A
B
Usually
chain
dependency
of
whoever
had
worked
on
it,
so
the
way
that
the
bot
will
assign
is,
if
you
have
history
and
the
file
may
funsies
here
is
if
you
do
get
move
across
time.
Sometimes
you
lose
that
history,
so
gleefully
I
don't
get
assigned
to
some
things,
an
API
machine
or
anymore
because
of
all
the
moves.
B
D
A
D
I
mean
it
wouldn't
be
I,
don't
think
it'd
be
too
hard
to
implement
your
preflight
check
using
the
YES
Network
package
directly
like
this
is
mostly
just
wrappers
to
make
the
the
previous
proxy
code
a
bit
cleaner,
so
I
can
probably
I
can
work
with
you.
Offline
see,
I
need
a
look
at
the
pre-flight
check
code
to
see
what
we
can,
what
we
can
duplicate
duplicate
if
we
have
to.
A
A
We
also
import
this
simple
package,
which
is
by
billing
I've
seen
this
around
on
the
web,
but
the
problem
is
that
we
already
have
the
ATM
machine
same
version
again,
so
we
should
use
that
I
think
so
actually,
I
sorry.
This
was
using
the
client.
Go,
not
the
API
machine
reversal.
It's
basically
that
it's,
basically
the
struct
that
when
you
build
a
binary
incorporate,
is
it
it
passes.
So
the
linker
flags
inside
this
structure
to
populate
you
know
the
the
version
of
your
binary
that
you're
currently
building
does
basically
change.
A
A
E
A
Basically,
the
problem
there
is
that
we
have
to
convince
signal
to
stop
importing
/t
in
this
very
medium
which
doesn't
make
any
sense.
The
you
know
some
tests
in
the
lot
integration
tests.
In
the
note
integration
package,
they
import
the
cue
medium
command-line,
basically
to
be
able
to
stretch
these
pre-flight
checks
for
docker
and
I
can
send
up
here
tomorrow
to
remove
all
the
profiles
we
do
for
the
doctor
API,
but
the
problem
is
that
we
are
going
to
break
this
signals.
A
A
B
What
point
does
it?
What
is
signals
official
state
still
support
for,
does
even
no
support
for
given
CRI,
because
in
the
ideal
world
we
kind
of
we
cut
out
a
lump
of
code,
if
we
just
sort
of
say
we're
all
going
to
use
container
D
we're
going
to
use
crowd
to
record
CRI
to
or
whatever
that
that
thing
is
called,
and
this
was
kind
of
like
the
in-state.
We
wanted
to
get
into
you
like
what
year
is
this
a
year
and
a
half
or
two
years
ago?
A
I,
don't
know
the
official
stance
at
the
moment
commander
cynic,
they
stopped
doing
doctor
validation,
which
was
a
signal
basically
I
recognized.
The
signals
do
not
think
the
doctor
is
a
primary.
She
right
at
this
point
in
you
know,
considering
don't
assume
as
well,
but
I
don't
think
they
see.
One
official
singles
here
I
that
we
should
yeah.
B
But
you
see
like
they
say
they
make
this
statement,
but
all
of
the
current
test
apparatus
uses
Google,
containerized
images,
and
you
can't
see
my
hand
close
here
in
the
McDermott
camera
Google
container
has
images
and
all
of
those
have
doctor
on
them.
So
basically,
the
entire
CI
signal
for
node
stuff
was
all
doctor
based
so
I'm
cool
with
them
saying
that.
But
we
need
to
actually
have
cross
coverage
of
Virna.
B
Make
this
real
right
either
that
or
we
make
me
default
all
of
our
images
that
we
baked
to
use
container
D
or
something
in
just
in
CRI
tool
or
something
which
would
be
another
route.
That
might
be.
That
might
be
a
reasonable
thing
for
us
to
do
as
a
separate
work
item
that
could
be
considered
higher
priority.
A
Yeah
I
don't
know
how
to
respond,
because
I
also
know
that
GE
has
docker
by
default.
Dr.
shim
is
going
away
eventually,
but
maybe
the
the
dr.
shim
going
away
is
going
to
be
bought
by
you
know,
G
key,
so
I
don't
know.
If
the
we
can
remove
this
signal
or
though
I
mean
Tokra
skills
to
say
in
general,
cooperate
is
I,
don't
see
it
going
away
anytime
soon,
so
I
don't
even
know
who
is
the
owner?
Who
is
the
person
I
should
talk
to
about?
A
E
A
A
This
is
the
occasion
where
we
do
the
Q,
medium
validator
stuff,
so
slash
test
slash
into
an
on
system.
This
is
where
we
import
/
CMD,
cube
ATM
for
the
sake
of
having
the
validators
and
I
I'm,
starting
to
think
that
we
maybe
as
a
short-term
item
for
the
sake
of
decoupling
from
this
in
eBay
again,
maybe
we
should
move
all
the
functions
inside
the
signal
add
stuff
inside
test,
slash
into
a
node
and
simply
stop
doing
these
API
type
checks
for
cube,
ATM,
it's
low,
it's
not
a
complicated
operation.
B
That's
why
I
said
like
we
should
really
be
relying
on
cry:
CRI
tool
right
in
the
Sierra
to
import
library.
That's
it
was
built
to
abstract
this
problem
away
and
it
to
work
across
whatever.
Why
don't
we
open
an
issue
specifically
for
this
one
in
in
deal
with
I
know
we
had
one
a
while
ago,
I
I
forgot,
which
one
it
is,
but
this
this
one
is
a
little
bit
more
30
yeah.
A
What
did
you
order?
The
validate
those
inside
the
you
know
the
slush
test
directory
at
some
point,
we
realized
that
Ola
cube
ATM
is
the
user
of
these,
so
he
moved
them
to
keep
ATM
at
some
point.
Signaled
said:
hey.
We
need
these
validators,
so
they
started
importing
/team
distally
medium
again,
it's
a
complete
myth,
I'm
going
to
create
an
issue
after
that.
A
Okay,
carry
file
the
so
many
scales
that
we
wanted
to
stop
importing
Caravelle
for
a
hundred
lines
of
code.
The
basically
the
maintainer
responded.
What
we
should
continue
to
import
car
caddy,
but
he's
going
to
refactor
the
package
structure
deems
help.
There.
I
also
had
some
questions
through
the
CSF
lawyer.
We
have
four
kubernetes
and
he
said
that
if
we
decide
we
can
copy
the
file
and
include
a
bunch
of
copyrights,
it's
it
is
doable,
but
the
maintainer
basically
said:
hey
I
am
going
to
help
you,
so
we
almost
resolved
here
for
this
particular
problem.
A
A
The
packet
we
covered,
the
packaged
stuff
recovered
some
of
the
vendor
stuff
and
the
rest
is
a
by
machinery.
Plant
go
and
know
llamo
and
que
walk.
So
that's
pretty
much
it.
This
was
more
of
a
walkthrough
and
we
didn't
get
comments
from
there.
You
know
the
public
types
API
Machaerus
toughs,
but
if
anybody
has
any
other
comments,
we
can
continue
this
discussion.