►
From YouTube: kubeadm office hours 2019-07-17
A
A
Basically,
this
of
x-cubed
iam
through
and
kind
I
guess
so
recently
a
conformance
test
was
promoted.
Sorry,
a
test
in
the
suite
was
promoted
to
conformance,
but
now
it
requires
the
metric
server
as
a
dependency.
So
basically
the
horizontal
pata
auto-scaling
tomato
sending
has
to
report
sees
the
errors.
It
has
to
a
metric
system
and
it's
technically
portable
to
any
metric
system.
But
apparently
the
test
requires
the
metric
server.
A
C
C
A
Yeah,
so
if
more
people
want
to
comment
here,
please
do
Clayton
on
slack
mention
that
he
expects
a
deployment
of
communities
to
always
in
South,
America
server.
I
think
this
is
this
might
be
true
for
products
in
a
corporate
environment,
but
for
users,
customers
that
want
to
be
conformed.
That
don't
think
this
is
always
the
case.
D
A
Not
when
the
setting
it's
workable,
so
the
system
is
going
to
detect
for
you
that
you
have
another
metric
system
and
it's
going
to
be
conformant,
but
this
particular
test
is
using
the
Masari
I,
don't
understand
exactly
the
details
I
just
found
about
this
today.
So
maybe
these
are
questions
that
we
should
be
asking
too
so.
C
I've
got
long-standing
opinions.
That
versions
of
the
API
is
that
we
created
an
eternally
we're
done
under
duress
and
that
there
are
better
ways
to
even
do
auto
scaling
than
using
the
eight
using
the
metrics
convergent,
using
the
tools
that
were
built
in
the
main
API,
so
that
we
never
had
a
discussion
because
HPA
was
was
in
alpha
forever.
C
Right,
like
it
was
an
alpha
and
beta
for
such
a
long
time
that
we
never
had
the
discussion
about
whether
or
not
this
should
be
a
v1
I
think
was
promoted
to
a
v2,
be
v1
by
people
in
absentia
and
I
need
to
go
to
conformant.
I
need
to
go
to
I
miss
the
conformance
meeting
this
week,
which
I
now
regret,
because
it
was
the
day-long
meeting
and
now
I
need
to
follow
up
with
folks
on
this
I
absolutely
think
it
should
be
recruited.
C
A
C
Don't
know
if
we
can
do
that
in
good
nature
right
now,
because
the
it's
it's
it's
a
long
story
that
require
too
much
background.
It
was
part
because
the
year
there
was
originally
a
deployment
of
hipster
which
had
that
separately
and
then
they
they
they
scooped
out
the
functionality
into
a
metric
server,
because
there
are
some
of
the
implementations
and
metrics
that
they
need
an
abstraction
layer
behind
that
right.
E
Okay,
I
want
to
make
a
proposal
for
making
meeting
mark
this
minute
parties.
You
can
play
me
because
discussion
gets
often
goes
in
details,
so
I
propose
that
we
try
to
be
a
little
bit
more
discipline
attend
and
we
organized
the
meeting
in
three
moments.
The
first
moment
is
to
welcome
new
contributor,
which
is
always
nice.
A
We
should
do
things
like
that
to
make
to
move
the
project
forward
faster,
for
instance,
last
time
we
didn't
make
any
good
decisions
in
terms
of
the
implementation
for
the
wreck
of
the
coaster.
We
only
make
made
some
decision
about
decisions
about
documenting
stuff,
but
yeah.
This
part
is
very
tricky:
the
first
half
you're
proposing
but
I'm
overall
and
+1.
For
this
and
I.
We
always
forget
to
welcome
new
people.
So
maybe
after
we
discuss
this,
we
can
have
this
opportunity.
E
Okay,
opinions.
E
A
G
G
I
should
be
interested
in
qadian
for
a
long
time,
but
my
kind
of
client,
where
work
hasn't
allowed
me
to
regularly
joint
calls
or
really
contribute,
but
recently
I've
been
working
a
lot
more.
The
customer
around
some
extra
pieces.
You
might
have
noticed
a
couple
of
issues,
I
raised
around
at
CD
upgrades
and
so
I
think
I'm
gonna
be
a
bit
more
involved
with
its
ETA
diem.
So
I'm
kind
of
interested
in
the
splitter
out
of
SCD
work
from
cube
ATM
to
ats-v
ATM
and
how
that
shakes
out
yeah.
A
E
Move
quickly
to
the
next
topic,
we
have
the
reason
we
are
in
flight
and
for
changing
the
activate
procedure
for
a
Shea
cluster,
and
this
PR,
who
was
the
follow-up
of
a
discussion,
started
by
a
file
and
basically
we
have
to
change
the
order
from
the
current
order.
Where
we
do,
we
have
to
grade
a
control,
plane
node,
including
the
coup
ballot,
and
then
we
moved
the
second
counterplay
node
a
button
as
soon
as
we
have
very
the
first
controller,
the
first
cupola
we
have
your
eating.
E
A
Yes,
basically,
I
set
the
PR
already
like
last
night
collagen
to
the
original
issue
as
well.
It's
it's
definitely
possible
from
to
hunt
to
something
bad
happen,
but
my
question
there
was
like,
if
you
have
multiple
control
planes
and
basically
RFI
saying
that
one
of
the
couplets,
if
it's
newer
than
the
API
server
in
the
cluster
and
if
they
are
not
on
the
same
notes,
I
think
basically
I,
don't
think
anything
else.
Anything
bad
is
going
to
happen
because
the
if
you
connect
your
workers
to
a
control,
plane
node.
A
E
D
C
E
On
the
issue
that
there
is
a
a
good
comment
from
the
traffic
okay,
this
one,
if
you
look
what
we
are
suggesting
now
team-
is
that
you,
you
first
upgrade
the
the
contravening
zero
and
you
agreed
both
okay
server
and
cooperate.
What
are
you
are
in
this
condition?
What
can
happen
that
they
could?
The
newer
kubilay
today
115
can
told
witha
one
14.1
api
server,
and
this
liberates
the
virtual
school.
C
So,
typically,
the
the
order
of
operations
was
control,
plane,
upgrade
first
and
then
goodness
second
right
and
that
would
even
include
the
hosting
coop
in
itself
and
because
of
the
couplet
upgrade
is,
is
rpm
/dev
upgrade?
Yes,
so
it's
not
even
part
of
the
upgrade
workflow.
It's
a
separate
step
triple.
E
A
Of
other
points
that
I
added
at
the
bottom
is
that
is
so
discussions
with
between
journal
of
Liberty
and
Joe
bits
from
Google
that
maybe
in
the
future
the
couplet
will
be
X
plus
1
compared
to
the
API
server.
So
maybe
the
secure
is
going
to
change
and
I.
Don't
have
the
link
I
forgot,
where
I
read
this
and
also
yesterday,
I
tested
a
couplet
that
is
newer
than
the
API
server
and
I
didn't
see
any
failures,
but
I
I
was
seeing
failures
back
in
the
1:14
cycle
when
something
changed
and
there's.
C
No
strict
guarantees
like
this
is
Jordan's
got
a
document
that
he
wrote
up
about
version
skew
policy
that
outlines
the
detail,
because
that's
the
policy
that
we
originally
have
there's
no
guarantees.
We
have
on
API
versioning
whatsoever
for
newer
versions
of
colet's
to
be
able
to
talk
shouldn't
some
VPS
or
newer
versions
or
triplets
to
talk
to
her
there's
no
doubt
great
conversion,
API
types,
if.
C
A
I
Hi
guys
just
a
few
quick
comments.
The
feature
gate
are
landed
for
dual
stack,
I'm
working
on
a
new
PR
which
will
basically
propagate
so
if
the
future
gate
for
dual
stack
is
set
at
the
cluster
configuration
level
will
propagate
it
down
to
the
individual
components
in
kubernetes,
so
I
have
that
out
shortly.
A
Yes,
so
basically,
this
is
something
that
I
discussed
with
you
dare
to
create
a
list
of
tasks
or
ask
first
enable
the
feature.
Gate
next
step
is
to
enable
the
feature
gate
to
be
passed
to
the
corporates.
I
mean
here,
we
can
possibly
split
I,
guess
we,
it
has
to
be
a
part
of
the
same
peer
to
enable
you
know
the
list
of
sitters
to
be
passed
to
the
components,
because
currently
the
coastal
city
is
a
single
string
upstream
is
changing
to
be
a
list
of
strings.
A
One
of
them
is
ID
before
one
of
them
can
be
ipv6,
and
the
last
step
was
to
enable
tests.
Well,
I
have
a
question
here
much
given.
We
are
going
to
release
this
as
alpha
in
116.
Hopefully,
how
does
this
relate
to
the
future
tasks?
Rd
is
going
to
be
better,
for
instance,
this
is
a
pretty
important
field
in
the
API
server.
A
I
That's
a
good
question,
so
I
I
think
the
current
state
is
that
you
had
a
very
sort
of
very
basic
PR
that
landed.
That
includes,
like
the
cluster
cider
change,
there's
a
series
of
PRS
that
are
going
on
by
different
people
who
are
gonna
bring
in
these
additional
changes,
whether
or
not
they
actually
hit
alpha
I'm,
not
entirely
sure
in
1.16,
but
yeah.
Eventually,
all
this
stuff
will
go
from
alpha
to
beta
status,
but
that's
very
much
in
the
future.
I
couldn't
really
really
comment
on
that
now.
A
A
I
I
think
it's
a
long
ways.
I
mean
it
depends
what
you
call
working
right.
For
example,
today
you
have
something
working
there
like
pods,
get
two
IP
addresses
and
they
can
actually
play
each
other.
If
you
set
up
your
cluster
properly,
and
so
it
really
depends.
What's
working
and
what's
not,
services
are
not
working
today
and
that
stuff
will
trickle
in.
I
You
know
so
anyone
who
actually
turns
on
this
feature
gate
really
needs
to
know
what
they're
doing
is
what
I
would
say.
I
and
it's
not
going
to
be,
will
be
working
for
a
long
time
and
fully
working
I
mean,
like
you
know,
fully
test
date.
Everything
is
going
going
grade.
You
can
difference,
try
different
configurations.
A
A
Good
as
well
yeah,
okay,
okay,
we
can
do
that
so
next,
so
this
is
somebody
brought
this
discussion.
It
was
in
the
VMware
slack,
so
basically,
they
had
a
problem
where
they
create
a
single
node
cluster
with
sorry
with
a
single
control
plane.
They
join
while
workers
and
they
they
decided
to
create
a
Haq
Sara
out
of
this
single
node
question.
A
But
to
begin
with,
there
were
not
using
the
control,
control,
pain,
endpoint,
and
this
means
that
you
don't
have
a
certificate
Verisign
to
use
a
control
plane
input
for
the
serving
that's
that's
the
service
sent
for
the
API
server.
The
basically.
My
proposal
here
is
to
created
a
section
of
the
create
a
single
node
cluster,
a
single
control,
a
node
Costa
Rica
medium.
The
section
of
the
document
can
contain
a
short
guide.
E
Agree
he
on
the
idea
of
writing
something
in
the
single
control
plane
document
in
order
to
make
his
make
possible
the
transition
to
multi
master
control,
plane,
I,
don't
think
that
we
will
not
tell
we
are
required
to
go
much
detail
so
something
like
an
oath.
That
is
plain.
It
is
highly
recommended
to
assign
Yanis
here,
Seanie
control
player
or
a
DNS
name
to
the
controller,
to
each
permitted
class.
That
can
suffice.
A
Basically,
this
is
solution.
Right
now
is
because
not
everybody
is
running
a
DNS
server
that
has
cname
support
tremendous
any
I.
Actually,
I
was
kind
of
confused
because
Boone,
who
did
a
lot
of
changes
in
the
newer
versions
and
I,
don't
know
how
to
set
up
cname
up
unless
I
asked
our
custom
DNS
server,
but
I'm
not
going
to
go
into
detail
since
she
name
and
cinema
Khorsandi
records
I'm
just
gonna
mention
them,
but
for
any
hosts.
A
What
you
can
do
is
you
can
create
a
hostname
that
maps
to
the
know
the
API
address
from
the
node
and
then,
when
you
pass
control
plane
a
point
to
be
discussed
instead
of
the
API
IP
address
the
joining
nodes,
if
they
still,
if
they
also
have
this
in
the
hosts,
ADC
hosts,
they
can
join
the
cluster,
and
eventually
you
might
decide
to
map
this
DNS.
Sorry
discussed
name
to
DNS
record
that
you
know
you
cannot
have
an
alias
or
you
can
map
it
to
a
different
IP.
E
E
Yeah
I
totally
agree
is
just
my
point
was
only
to
keep
it
as
simple
as
possible.
Yes,
okay,
okay,
I'm.
Next,
so
at
the
end
of
the
man
who
had
the
first
deadline
of
this
cycle
and
I
I
tried
to
make
up
my
mind,
what
is
the
status
of
of
the
current
priorities,
and
especially,
we
forgot
the
deadline
so
adding
all
day
that
that
requires
web.
All
the
cap
and
we'll
day
announcement
issue
in
place
and.
E
E
F
E
E
E
E
E
E
E
E
E
B
A
Ticket
for
and
I
will
check
yeah.
We
don't
need
a
cap
for
this,
so
my
my
thinking
is:
if
something
is
complicated
and
meet
needs
a
lot
of
discussion,
then
it
needs
a
cap.
Otherwise,
if
we
are
fixing,
potentially
you
know
10
lines
of
code
to
make
something
better.
I,
don't
think
it
needs
a
cap.
Yes,
I
am
is
just.
E
E
E
C
A
There
is
a
PR
employed
by
a
microscope
or,
but
it
was
delayed
because
Microsoft
has
come
for
licensing
problems
with
some
of
their
PowerShell
scripts
and
I
was
in
the
SiC
windows
meeting
yesterday
there
was
a
basically
a
short
comment
requesting
review
from
Patrick
Blanc.
So
it's
basically
both
on
P
point
prev
of
this
point,
and
this
is
for
the
Alpha.
So
currently
we
are
in
pre-alpha
cube.
A
DM
is
ready,
but
the
scripts
that
rabbika
medium
for
Windows
Server
dream.
This
is
the
status
of
it.
A
A
E
E
K
E
A
A
E
A
A
E
A
E
A
A
A
So
this
this
PR,
we
hope
we
put
a
hold
on
that.
Basically,
the
contributor
I
wanted
to
mount
certain
certificate
directories
into
the
schedule
report,
but
me,
and
the
dear
basically
explain
to
him
that
it's
not
really
needed.
Kapadia
already
supports
a
mechanic
for
that
using
extra
volumes
and
adding
this
extra
certain
keys
is
not
really
required
from
from
a
security
standpoint,
so
this
PR
we're
probably
not
going
to
budget.
A
C
Yes,
because
we
actually
had
that
capability
for
a
long
time,
right,
I
think
we
just
somehow
regressed
I,
don't
I,
don't
necessarily
know
if
it
mean
just
patching.
The
simple
scenarios
that
we
missed
seems
like
legitimate
fix.
I,
don't
know
if
changing
behavior
is
is
what
we
want
for
this,
because
we
always
had
the
capability
of
comedian
to
fetch
from
the
internet.
The
latest
version
and
that's
kind
of
a
user
expectation
down
so
yeah.
A
C
This
was
very
early
in
the
COBE
DM
days.
I
think
it
had
to
do
with
the
fact
that
you
didn't
want
to
have
to
go
through
the
entire
installation
procedure,
especially
for
patch
renditions,
to
get
the
latest
patch
version.
So,
if
I
have
conveyed
him
locally
installed
in
my
environment
and
a
new
patch
version
comes
out,
I
have
the
ability
to
you,
deploy
an
update,
fairly
seamlessly
without
having
to
do
a
whole
reinstall
of
the
application
on
the
host
machine.
A
Yeah,
because
if
we
so
if
the
user
starts
with
an
O
code,
qadian
binary,
it's
going
to
definitely
install
using
this
mechanic
is
going
to
install
a
new
version
of
the
control
plane.
The
latest
patch
stable
patch,
which
is
fine,
of
course,
but
they
guess
they
are
stuck
with
this
version.
So
the
next
patch
version
they
are
not
going
to
install
it
because
the
question
is
already
up.
A
You
know
I'm
not
arguing
about
this.
It's
just
that
it
created
such
a
mess
with
air-gapped
scenarios
and
everything
you
know.
So
we
are
still
battling
the
decisions.
Whether
we
should
you
know,
handle
five
hundred
HTTP
responses
as
something
that
should
be
tolerable
and
things
like
that.
I
agree
that
it's
definitely
a
breaking
change.
If
we,
if
we
enable
the
the
version
of
the
client
by
the
by
default,
it
just
I.
C
Don't
have
strong
opinions,
it's
more
legacy
than
anything
else.
I
I,
don't
mind
like
having
an
optional
fetch
latest
or
something
like
that
that
a
person
can
opt
in
to
you
and
then
the
default
is
always
the
client
I'm
not
opposed
to
that
I
think
the
people
you'd
have
to
poke
the
most
on
I
think
would
be
Alexander
and
Lucas
to
verify.
A
A
A
As
I
said
earlier,
the
fields
some
of
the
fields
in
key
proxy
now
are
a
slice
of
strings
instead
of
like
a
regular
strings,
so
we
I
think
we
have
to
think
about.
This
is
what
this
does
not
have
a
milestone
yet,
but
maybe
we
should
start
investigating,
like.
Are
we
okay
to
transition
to
V
1
alpha
2
in
our
queue
proxy
deployments?
Rubidium
I.
A
D
D
So
the
idea
was
to
extract
these
into
the
YouTube
repo
and
once
the
once,
these
are
used
out
of
three
we
can
cancel
like
the
idea
is
to
see
with
zigge
Network
if
they
can
reuse
this
function.
These
functions
and
implement
like
implement
these
function
under
a
staging
repo,
I
think
the
proxy
one,
although
for
now
they're
just
using
it
for
the
api's
like
in
the
functions
there
and
moving
and
using
the
things
that
we
move
to
Utah,
then
us
consider
these
functions
of
under
proxy
in
TBD.
The
goal
is
that
they
should
maintain.
D
A
Yeah
so
I
think
I
made
a
a
bit
of
a
rant
on
slack
in
the
TLDR
of
my
rant
is
that
it
feels
to
me,
like
we
shouldn't,
be
checking
for
a
AP,
I--,
p
vs
on
the
nodes
at
all,
and
it
should
be
the
responsibility
of
the
proxies
to
do
this
for
us,
because
we
are
in
T
cube
ADM.
We
are
integrating
the
part
of
the
backend
of
proxy
for
the
sake
of
performing
a
validation
that
you
know,
it
should
be
the
responsibility
of
the
other
component,
and
this
is
basically
I
made.
This
comment.
D
A
D
D
A
D
A
That's
unclear
that's
up
to
Sigma
Network
to
decide
basically
to
man
deciding
the
purpose
of
the
repository
was
true
cost
not
only
the
config
but
the
other
stuff
as
well,
but
you
should
really
ask
them.
The
same
goes
for
the
kubernetes
/
:,
the
repository
under
staging.
Can
we
host
couplet
utyos'
in
there?
Can
we
host
the
you
know
the
get
hostname
function
that
Ed
was
working
on
fixing
like
what
is
the
state
of
this
reporters
I'll.
A
Are
small
brokers
that
pretty
mature,
prevent
qadian
from
moving
out
without
the
decision
at
this
point?
I
just
proposed
that
we
should
remove
this
check
completely
from
cube,
ATM
and
other.
You
know
other
release,
lot
about
it
and
say
it's
a
complication,
dependence,
complication
and
maybe
one
day
we
can
introduce
it
back
in
another
form.
That
was
my
proposal.
I.