►
From YouTube: kubeadm office hours 2019-12-11
A
Hello,
everybody-
this
is
the
sequence
of
psychic
medium--
office
hours
today
is
the
11th
of
December
2019,
and
this
is
before
Christmas,
but
we
decided
to
have
our
planning
session
for
the
next
cycle
180
and
we
already
have
a
document
that
I
filled
with
some
items.
So,
let's
begin
by
sharing
the
screen.
A
No
major
hiccups
release
loads
needed
some
tweaks,
so
we
we
had
to
edit
some
things
in
the
restalls,
because
some
sort
of
a
mixture
happened
between
action
required
items
and
items
that
are
without
action
required,
but
we
managed
to
fix
that
also
pieces
that
George
alacrán
was
appointed
for
release.
Lead
of
1:18
George
was
is
helping
about
insig
testing
and
testing
areas.
A
Takes
Merrick,
oh
yeah,
we
seem
to
have
lost
agenda.
Somebody
added
it
back
all
right
now,
going
to
the
planning
for
today.
So
a
quick
legend
explaining
what
the
priorities
means
here.
The
priorities
mean
here
is
that
we
basically
have
a
list
of
items
we're
going
to
discuss
each
item
and
at
the
end
of
the
discussion
for
this
particular
item,
we
have
to
assign
a
priority
for
it.
So
p0
is
the
highest
priority
before
something
like
needs
evidence,
it's
the
lowest
priority
and
also
when
working
with
something
we
have
to
assign
people
to
them.
A
Some
items
we
transfer
from
the
previous
cycle
and
they
already
have
assigned
people
to
them,
but
if
more
people
want
to
help,
there
is
opportunity
to
add
you
here
instead
of
the
question
mark
alright.
So
the
first
item
of
the
list
is
something
that
we
promised
for
the
secret
texture
and
that's
moving
Hubei
the
amount
of
the
kubernetes
core.
It
is
repository
to
the
cube
ADM
repository,
and
we
have
tracking
issue
for
this.
A
It's
here
and
also
I
managed
to
compile
a
pretty
big
Google
Documents
about
all
the
potential
problems
we
are
going
to
face
with
this,
and
we
we
got
a
lot
of
comments
from
people.
Thank
you,
everybody
for
commenting
now,
there's
a
there's,
a
question
whether
we
should
continue
commenting
here,
solidify
our
decisions
and
create
a
kept
or
we
should
potentially
keep.
A
C
E
A
A
All
right-
and
my
proposal
here
is
that,
based
on
our
pretty
much
I
know
how
to
say
this,
like
we
said
to
seek
architecture
that
we
are
going
to
do
this
in
118.
So
there
are
some
things
to
discuss
with
other
six
like
secretly
sick
dogs,
but
on
our
side.
Maybe
this
is
the
highest
priority
for
one
item
and
I
was
proposing
p0
here,
I'm.
A
F
F
G
H
F
G
A
A
G
G
The
reason
I'm
working
on
the
PR
in
concert
with
the
cap
is
so
that,
like
the
ideas
are
fully
fleshed
out
in
the
cap,
because
it's
a
rather
large
feature
that
has
many
UX
touch
points
throughout
all
of
Cove
idioms
lifecycle,
so
but
yeah
I'm
prioritizing
getting
back
to
a
lot
of
these
questions
and
clarifying
with
docks
in
the
cap.
I.
A
See
yeah
so
I
guess
we
have
to
have
all
the
questions
answered
and
probably
me
Fabrice.
You
have
to
go
again
at
the
those
diagrams
that
we
created
for
all
the
interaction
with
cue
medium
between
the
other
ancillary
medium
and
we
I
guess
we
have
to
say
that
this
is
the
kept
is
at
this
point,
implementable.
G
D
G
Yeah
I
do
agree
with
the
comments
on
feature
parity.
We
need
alternate
or
equivalent
mechanisms
for
all
the
ways
that
we
currently
treat
the
admins.
A
G
A
I
A
Sorry
we
block
to
the
PR,
but
the
it
was
a
very
clear
like
we
requested
this
last-minute
change
streaming
versus
long
streaming
and
the
structures,
in
my
opinion
personally
got
a
bit
weird
yeah.
A
I
J
A
B
A
A
B
I
made
a
couple:
I
made
a
couple
of
kind
of
I
made
some
stuff
that
I
actually
want
to
go
over
there
with
you.
So
I
made
a
couple
of
PRS
that
kind
of
do
some
of
the
cleanup
that's
needed
to
implement
the
phases
and
I'd
actually
love
to
do
like
a
you
know
an
hour
some
time
with
you
to
go
over
them,
because
I
I
know
that
you've
done
this
before,
but
yeah
I
I'm
moving
forward
with
it
I
hope
to
have
the
first
of
a
set
of
PRS
in
by
the
end
of
December.
A
Yeah,
so,
okay,
thanks
going
back
to
retrospective,
like
for
the
in
it
enjoy
phases,
we
were
able
to
assign
multiple
people
to
walk
in
on
individual
phases
was
the
the
runner
the
phase
runner
was
already
got
into
a
working
state
for
apply.
Basically,
we
had
multiple
people
work
on
the
phases.
So
if
you
are
able
to
send
this
primary
PR,
we
can
then
start
delegating
the
work
to
others.
A
G
A
A
So
no
see
this
is
the
V
1
beta
3
item.
We
have
an
issue
actually
I
create
a
like
filter
here
for
all
the
API
changes
we
have
and
the
most
recent
one
was
worse,
basically
under
ghosting
recommended
that
we
should
have
image
full
policy
defined
in
the
cube
ATM
configuration
file,
because
some
people
want
to
always
pull
the
images.
Some
people
want
to
pull
only
if
missing
work
early,
for
instance
or
maybe
never
so.
Basically,
he
suggested
we
should
add,
like
the
image
full
policy
for
in
our
config.
A
F
So
I
think
that
we
should
reconsider,
adding
give
you
a
three-in-one
18,
but
we
also
need
to
consider
object,
meta
and,
like
the
rest
of
the
stuff
that
we
actually
figured
out
is
probably
not
really
needed
and,
for
example,
if
I
can
actually
go
out
without
object
matter.
It
will
be
much
more
like
sensible,
at
least
from
cube,
ATMs
perspective,
and
if
we
are
actually
left
only
with
image
policy
and
with
removal
of
the
huge
hypercube
image
boolean,
then
it's
probably
not
worth
to
do
a
whole
new
version.
A
A
A
A
A
Yes,
so
the
bots
automatically
assign
folks,
like
you
seen
committed
here,
said
the
cue
is
also
in
the
list
and
also
the
QA
I
added
him
here,
because
he's
already
he's
already
helping
in
this
effort
for
unit
tests
at
least,
but
do
you
Felicia?
Are
you
looking
for
more
people
to
be
like
dedicated
reviewers
in
this
effort.
A
A
question
I
have
for
Tim
who's
on.
The
call
me
for
bridge
also
managed
to
discuss
this
topic
quickly
and
I
was
wondering
whether
we
can
mark
the
kept
implementable.
If
we
can
assume
that
we
can
go
with
a
primary
stage
of
having
basic,
not
not
a
declarative
approach
at
the
beginning,
but
then
transition
to
the
clarity
of
approach.
A
A
H
D
D
A
H
Say
yes,
because
they're
gonna
be
locked
step
with
capabilities,
if
you
eventually
maybe
out
of
tree
but
I,
think
until
we
get
into
GA
or
the
the
decoupling
is
super
clear
with
regards
to
the
API,
because
don't
forget,
our
API
is
still
beta
right
and
you
potentially
could
have
impactful
changes
on
the
API.
What's
the
contract
for
the
API
is
is
clear
that
I
think
it
makes
sense
if
people
decide
that
that's
what
they
want
to
do,
I
think
for
the
time
being,
because
there
would
be
a
higher
degree
of
churn
and
they
will.
A
A
A
We
we've
been
dragging
this
for
a
while,
so
something
that
sig
windows
asked
me
personally
to
help
with
is
to
get
the
work
or
no
upgrades
working
and
I.
Think
it's
a
matter
of
calling
the
cube
ATM
upgrade
mode
command.
At
this
point.
It's
not
anything
special,
but
I
just
need
to
find
the
time
to
do
this
and
from
there
we're
kind
of
blocked.
We
we
want
to
support
this
feature
at
some
point
properly,
it's
still
in
alpha,
but
we
also
blocked
on
end-to-end
tests
and
for
end-to-end
tests.
A
We
need
people
to
sign
up
for
creating
the
infrastructure
and
also
like
getting
the
deploy
or
for
tests
working,
and
this
is
not
something
that
is
on
us
technically,
like
a
trike
helping
out.
If
some
people
in
the
kondakov
want
to
help
in
this
effort
in
terms
of
knowledge
about
Windows
PowerShell
and
things
like
that,
I
can
add
you
to
the
list,
but
it's
for
a
strict
requirements.
This
is
something
on
me:
I.
A
D
A
A
So
this
is
such
a
sort
of
fabricated.
This
item,
because
we
like
we
accumulated
a
number
of
changes,
related
columns
that
we
want
to
make.
We
have
some
different
issues
locked
in
the
tracker,
and
this
is
the
same.
These
are
the
same
like
the
problems
here.
So
basically,
our
upgrades
correctly
stomped,
the
existing
user
modified
coordinates,
coordinates
deployment
object,
for
instance,
if
they
use
the
autoscaler
to
change
the
replica
count.
We
stomp
this,
we
so
somebody
from
VMware,
actually
I'm,
sorry
I,
think
Peter
I
forgot
the
name.
A
Sorry
he
said
appear
to
fix
this,
but
they
only
fixed
the
replica
problem
and
it
kind
of
is
becomes
a
question
like.
Should
we
respect
the
existent
deployment
because
you
correctly,
we
completely
replace
it
during
upgrades.
My
opinion
was
that
we
should
respect
it.
So
that's
like
one
of
the
problems
there
a
promise
that
the
replica
count
like
graphically
a
separate
document
Feliz
how?
How
should
we
handle
the
replica
count?
Should
we
scale
it
up
when
new
modes
joining
I
should
use?
A
demon
said
like
there
are
different
ways
to
approach
this
problem.
A
The
final
one
is
Q:
penis
is
becoming
a
maintenance
bulletin
for
us
with
all
the
migration
stuff
and
all
the
that
some
of
these
changes
who
want
to
do
right
now.
We
also
have
to
manage
cube
DNS
and
it
feels
feel
like
we
probably
could
medium
doesn't
have
a
DNS
user.
So
this
this
point
and
also
it's
possible
to
skip
the
album.
A
So
this
is
this
is
basically
related
to
the
work
that
mark
is
going
to
do
with
a
upgrade
apply.
If
they
skipped
the
I,
don't
they
can
do
whatever
they
want?
Taking
the
boy
customer
one
and
cube
DNS
could
be
this
custom
DNS
server
that
they
Depot
instead,
so
I
want
to
potentially
get
some
work
done,
and
some
of
these
items
in
118
so
I'm
sure
laughter,
I
want
to
help
in
this.
A
M
A
M
Yeah,
it's
it's
kind
of
tricky.
Of
course
we
know
that
if
someone
makes
changes
the
deployment,
do
we
see
that
when
we
upgrade
a
lot
of
things
are
replaced
by
during
the
upgrade?
But
if
we
need
to
add
some
things
to
the
new
deployment
like
we
did
a
couple
of
releases
ago
for
the
readiness
providing,
then
we
need
to
support
such
cases
as
well.
A
D
D
A
Yes,
this
was
my
proposal
to
marketers
deprecated
118,
which,
because
this
is
GA
functionality
is
going
to
take
three
cycles
but
minimum
for
us
through
the
site,
if
you
want
to
remove
it,
but
I
wanted
to
deprecated
118,
because
it's
just
difficult
to
manage
these
two
options.
To
this
point,
given
to
my
knowledge
for
adoptions,
is
not
used
at
all.
At
this
point,
I.
F
H
I
would
like
to
do
almost
from
the
survey
through
the
whole
sake,
because
there's
broader
questions
beyond
could
be
to
him
that
I
think
are
really
important.
There's
a
lot
of
use
cases
for
cluster
API,
as
well
as
I,
want
to
understand
a
lot
of
friction
points
that
exist
for
configuration
for
another
providers,
so,
like
I,
think
if
somebody
want
I'll
happily
work
on
this
with
somebody
else,
but
I
think
I
want
I,
want
broader
feedback
across
the
stage
where
possible
here
so.
H
A
L
D
A
D
Okay,
so
the
story
is
that
in
the
inter
cluster
status,
we
we
are
maintaining
the
list
of
the
often
points
and
we
use
these
for
joining
when
joining
us
city.
But
what?
If
a
node
is
removed
without
doing
the
rosette
the
the
list,
the
list
of
endpoints
gets
stale
and
this
creates
problem
when,
when
you
join
a
new.
D
Anonymity,
3
node
and
this
cap
is
basically
proposed
tool
to
remove
the
list
and
instead
of
looking
at
the
list,
we
look
at
the
Adept
static
pod
that
this
is
deep
water
that
are
in
the
clusters.
So
we,
instead
of
maintaining
at
least
we
do
cluster,
is
passional
of
the
cluster
and
this
should
work.
But
there
are
some
comments
that
if
we
are
going
to
remove
the
cluster
status,
we
are
going
to
break
user.
D
A
D
L
D
D
D
D
A
G
If
the
reality
is
that
there's
not
one
Kubla
config
that
can
be
applied
to
all
nodes
in
every
cluster,
so
like
most
or
some,
some
users
have
clusters
where
nodes
need
different
configurations
for
their
couplets
and
we'd
like
to
be
able
to
expose
some
mechanism
for
people
to
do
this
in
kopitiam.
Since
kuben
m
wants
to
manage
this
configuration
for
people,
or
at
least
have
an
answer
yeah
so.
H
So
we
could
have.
You
could
upload
multiple
accounts
for
different
good
lists
like
you
have
no
rules.
For
example,
I
haven't
known
true
love
like
certain
types
of
nodes
called
X
and
certain
types
in
those
talents
called
Y.
You
complete
them
if
those
different
pools,
so
this
pool
of
nodes
could
then
point
those
to
different
context,
but
it
might
have
been
lost
in
translation
somewhere
or
in
lost
somewhere
in
the
PR
process.
F
G
G
Yeah
the
contributor
who
prompted
this
conversation
from
working
component
standard
mentioned
some
of
the
couplet
flags
that
have
not
been
migrated
to
the
config
yet
and
that
surfaces
I
think
this
exact
same
issue.
And
then
you
said
you
were
speaking
about
it
from
a
coup
proxy
perspective.
I
believe
so
I
agree
and
would
be
happy
to
be
involved
on
mechanisms
that
don't
exist
yet
that
are
necessary
for
users
to
do
whatever
is
necessary
to
maintain
the
config
for
different
permutations.
F
G
B
D
Below
cotton,
no,
because,
basically
now
there
is
no
support
for
change
in
the
cluster,
but
if
we
want
to
support
a
change
the
cluster.
This
is
something
that
we
need:
yeah,
okay,.