►
From YouTube: 2020-03-11 - Cluster API Office Hours
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
B
A
Api
is
a
sub-project
of
sig
cluster
lifecycle.
This
meeting
is
being
recorded
and
we'll
post
it
to
our
YouTube,
the
cluster
life
cycle
playlist.
Once
it's
ready,
please
use
the
raise
hand
feature
in
zoom.
If
you'd
like
to
talk
about
something
and
I
will
call
on
you.
We
have
an
agenda
document
which
I'm
currently
sharing.
Please
add
your
name
to
the
attendee
list
and
if
you
have
any
agenda
items,
please
add
them
there.
A
So
I
will
do
what
we've
been
doing
for
the
past
few
meetings
and
just
pause
to
say,
welcome
to
anyone
who's
new,
and
if
this
is
your
first
time
joining
glad
to
have
you
and
if
it
is
your
first
time
and
you'd
like
to
introduce
yourself,
I
will
pause
for
a
minute
or
so
to
see.
If
anybody
wants
to
say
hi.
A
C
C
That's
in
last
night,
it
has
like
some
good
getting
started
and
like
kind
of
an
overview
of
the
changes
and
the
release,
notes
for
sure
I
guess
we
do
have
one
bug
around
upgrade
that
and
you
fixed
before,
so
we
probably
cut
our
zero
three
one
soonish
give
me
that
was
a
nullpointerexception.
But
again
thank
you.
Everyone
for
the
huge
contributions.
D
Working
on
a
blog
post,
talking
about
the
Alpha
3
release
and
it's
in
a
Google,
Doc
I
would
love
you
know
to
get
everyone's
eyes
on
it
to
get
feedback.
It's
not!
It's
not
done.
I'll
be
working
on
it
throughout
this
week.
I
need
to
add
some
some
more
snippets
and
and
examples,
and
things
like
that,
make
it
a
little
make
it
come
alive.
A
little
bit
more
and
I
also
need
to
cover
a
few
topics
that
I
haven't
yet
one
being
the
Machine
machine,
health
checks
and.
D
The
end-to-end
his
framework
and
yeah
I'm
gonna
reach
out
to
the
people
on
on
slack
with
questions
and
yeah.
If
you
have
you
know,
if
you
have
comments,
questions,
complaints
or
concerns,
please
you
know
you
leave
them
on
the
docker
or
ping
me,
and
this
is
this
is
going
to
be
wrapped
up
for
submission
next
Monday.
A
E
A
Already
sorry
yeah,
we
can
go
on
to
let
me
reshare
mine,
real,
quick
Jason.
Did
you
want
to
go
over
the
failure
domain?
Tweaks?
Yes,.
F
F
A
F
G
H
G
A
H
E
D
E
Cool,
so
what
I've
done
is
I
set
up
a
client
caster
based
off
of
a
cluster
API
cluster
based
off
of
North
three
I'm
using
AWS
the
Norfolk
one
I
see
one
I've
got
the
log
streaming
from
the
controller
up
here.
The
machine
controlling
help
tracking
set
I've
got
four
machines
running
one
control
plane
in
three
workers.
I've
got
some
the
nodes.
Obviously
here
represented
and
all
I'm
going
to
do
is
apply
a
machine
health
check
so
in
this
bottom
left
corner.
E
Here
is
what
that
machine
health
check
looks
like
we've
got
a
cluster
named
ops.
Ii
got
this
crime
max
I'm
healthy
label
selected
the
machines
they
should
select
and
the
conditions.
So
the
idea
with
the
Machine
health
check
is
that
if
any
of
these
conditions
are
matched
for
this
time
out
period,
then
the
Machine
should
be
considered
unhealthy,
and
at
that
point
the
Machine
health
check
induce
remediation,
typically
I'd
expect
you
to
set
it
longer
than
30
seconds,
but
demo
wanted
it
to
be
quick.
So
let
me
apply
that.
E
E
So
Kubler
obviously
stops
the
ping-pong
between
the
api
so
from
qutilat,
meaning
that
eventually
it's
going
to
go
unready
here
and
what
we
should
see
is
that
the
Machine
healthcheck
controller
will
notice
that
go
on
ready
and
then
start
saying,
there's
a
target
that
is
I'm
ready
and
that
it
can't
actually
do
anything
not
sure
how
long
it's
going
to
take,
hopefully
not
very
long.
What
I've
done
at
the
moment
is
I've
set
the
max
on
healthy
parameters,
temp
cent,
which
is
very,
very
low.
E
E
E
E
E
So
when
it
sees
it's
not
ready,
what
it's
going
to
do
is
it's
going
to
here
we
go
so
you
just
want
to
highlight
some
of
this
locked
line
here,
short-circuiting
really
nation.
It's
a
take
that
there
are
nodes
unhealthy,
but
because
the
maximum
I'm
healthy
allowed
is
10%.
That
is
not
going
to
do
any
remediation
so
to
allow
it
to
remediate.
E
E
A
A
D
Thank
ya,
I'm
curious,
so
does
max
unhealthy
soso
you
it
looked
like
you
increased
the
max
unhealthy
to
allow
to
unblock
remediation
right
so
so
is
is
max
unhealthy
sort
of
like
analogous
to
the
like
a
disruption
budget
is
that
yeah.
E
D
Interesting,
okay,
I
I,
guess
I
I
found
it
I
was
a
little
confused
because
I,
you
know
I.
So
in
this
case
right
we
had
three
three
nodes
right
and
so
max
unhealthy
was
initially
at
10%,
and
so,
when
one
node,
you
know
when
in
and
out
ready,
III,
truly
I
thought
or
I,
don't
know,
I
shouldn't
say
to
Allah
I
thought
that
okay,
now
we
have
34
here
you
know
33%
unhealthy
and
so
then
I
thought.
Oh
well,
you
know
who's
got
33%
unhealthy.
Then
that
succeeded,
that's
unhealthy!
D
So
now
it's
time
to
remediate
or
met.
You
know,
if
it's
you
know,
but
if
it
were
than
the
threat
threshold,
maybe
you
wouldn't
remediate
so
there's,
maybe
maybe
it's
just
the
I,
don't
know
if
anybody
else
had
a
similar
thought
or
if
there's
you
know,
if
you
know,
if
there's
a
if
it's
something
to
do
with
the
the
name,
you
know
unhealthy
kind
of
as
a
low
as
it
can
both
describe
it.
D
E
A
D
It's
maybe
something
maybe
using
the
the
term
disruption
might
might
help,
but
anyway
I'm
happy
to
take
it
offline.
Well,
the
demo
looks
great
I
I,
you
know
demo
God,
notwithstanding
I
think
this
is
really
cool.
I
know.
A
Oh
ready,
Vince
I
see
you
added
a
discussion
topic
yeah.
C
So
in
the
past,
with
it,
some
could
walk
through
to
one
like
kind
of
like
four
new
contributors.
We
also
recorded
the
user,
which
have
been
very
helpful
and
also
like
to
just
like
see
things
from
a
different
perspective,
Fowler
issues
and
things
like
that.
I
also
really
wanted
to
do
a
core
walkthrough
for
ghosts:
ricardo
BTO.
I
also
thought
about
like
adding
cube
idiom
control,
plane,
I'm
wondering
like,
is
there
anything
else
that
folks
would
like
to
like
into
a
Code
walkthrough
in
the
Kappa
code
base
and,
if
so
like?
C
If
there
are
any
volunteers,
do
so,
you
don't
have
to
be
super
familiar
with
the
code,
but
it
would
be
like
one
of
us
that
would
probably
be
there
that's
more
familiar
with
the
code.
The
idea
is
that,
like
people
seeing
the
code
with
different
perspective,
different
eyes
will
find
things
that
we
might
wanna
like
fix
or
change
and
improve.
H
Hello,
I
think
that's
a
great
idea.
I
mean
I
would
love
to
go
through
some
code,
maybe
for
the
whole
thing
there
is
this
one,
video
I
found
from
summit
from
two
team
I
think
and
there
was
the
internal
machine
versioning
and
what
is
what
they're
present.
So,
if,
if
we
could
do
something
similar
for
caster,
API
I
think
it
was
keeper
psyche,
I
thought
it
was
nicely
explained.
C
A
Okay,
Thank
You,
Vincent
puddle
all
right.
Next
up,
we
have
James
with
managed
cluster
support,
update
I.
I
So
I
came
into
the
meeting
last
week
and
just
kind
of
asked
around
where
we
were
standing
with
this
so
I
read
through
the
issues
and
I
also
talked
to
a
few
folks
here
on
the
Microsoft
side
and
somebody's
actually
put
together
a
prototype
of
this,
and
it
looks
like
we
could
probably
manage
without
having
to
make
any
adjustments
to
the
rough
work.
The
specs
and
so
I
think.
I
A
A
Nadir
are
you
here?
No,
let's
see
I
think
that.
A
We
probably
do
need
to
track
down
and
do
some
more
investigation
into
this,
given
that
it
seems
that
at
least
with
the
Covidien
control
plane,
we've
been
seeing
some
issues
with
joining
and
with
etsy
dbits.
So
for
right
now,
I'm
going
to
put
this
in
zero,
3x
and
I
know.
Jay
had
created
this
one
which
we'll
talk
about.
A
We
can
talk
about
it
now
about
trying
to
simulate
some
of
this
with
the
docker
provider
for
cluster
API
and
I
think
these
are
probably
interrelated,
so
I'm
gonna
put
them
in
the
milestone
and
for
right
now,
especially
since
he's
not
here.
I
signed,
both
of
them
to
Jay
and
I'll,
follow
up
with
Jay
offline.
A
G
I
this
was
originally
filed
in
caps
me
by
someone
who
was
trying
out
cap
C,
and
we
had
a
short
discussion
with
Jason
there,
and
he
said
it
might
be
good
to
bring
this
back
up
to
khaki
and
that
at
the
time
it
was
decided
that
there
shouldn't
be
a
common
label
implementation.
But
I
just
wanted
to
start
the
discussion
again.
A
There
was
a
an
old
kubernetes
issue
around
worker,
node
labeling
and
from
what
I
recollect
there
are
no
standard
predefined
labels
for
anything
outside
of
the
control
plane
and
it's
up
to
a
provider
or
an
implementer
or
systems
integrator
such
as
cluster
API
or
any
other
tooling,
to
decide
if
they
want
to
do
that.
So
I
think
it's
really
up
to
us
as
a
community
to
decide.
If
we
want
to
do
this,
Fabrizio
I
see
you
end
up,
have.
A
J
A
K
One
is
actually
sort
of
contentious
and
I
think
it
the
use
of
these
labels
and
I.
Think
we
as
I
recall,
we
said,
no
drill
is
not
gonna,
be
formalized
in
KK,
but,
for
example,
we
could
absolutely
define
it
in
cluster
by
cycle
rate.
Yes,
we
can
all
agree
something
we
can
all
do
it
they're.
Just
not
nobody
in
KK
great
I
would
like
I
would
love
to
see
us
to
find
that
to
be
honest
but
yeah.
It's
doesn't
us.
We
have
seven
in
this.
We
should
define
any
cusp
of
life
cycle.
Yes,.
A
A
So
this
one
is
thanks
to
some
work
that
Lauren
and
Vince,
and
others
worked
on
we're
now
able
to
update
for
DNS
as
part
of
the
cube
idiom
control,
plane
upgrades,
which
includes
migrating
the
core
DNS
configuration
file
if
there
are
any
changes
from
one
chord,
E
and
s,
release
to
another.
As
part
of
doing
that,
we
have
taken
a
dependency
on
this
core
DNS
migration
project
and
it
was
essentially
a
stopgap
measure
until
we
come
up
with
a
better
plan.
L
I'm
all
about
finding
a
better
plan,
especially
using
their
operator,
they
have
an
operator
now
that
we
should
be
talking
with
them
about
to
do.
The
lexical
min
has
been
according
this
and
that'll
be
integrated
with
this
separate
ions
project.
I'm
gonna
file
budget
issues
around
the
stuff,
so
my
my
goal
would
be
that
we
should.
This
is
a
specific
area
where
I
think
we
should
start
to
align
into
some
projects
and
communicate
more
broadly
across
these
areas,
because
it
doesn't
make
sense
for
individual
folks
to
be
doing
different
things
across
the
different
sub
projects.
L
A
B
A
B
Basically,
it's
currently
already
works,
but
it's
it's
a
container,
sorry
controller,
runtime
v1,
so
he's
currently
updating
it
to
v2
and
after
this
is
out,
we
can
start
looking
at
the
operator
now
the
others.
This
is
like
a
separate
topic.
Do
you
want
to
use
it
with
the
alimony
seller,
but
yeah
I
just
want
to
say
this
that
the
operator
is
pretty
much
there.
A
All
right
next
up
is
see
if
we
can
remove
install
AMD,
we
have
a
basically.
If
you
look
at
the
QuickStart,
the
QuickStart
says
you
need
to
have
cute
control,
and
then
you
need
a
bootstrap
cluster.
Then
there's
some
provider
prereqs.
Then
you
get
cluster
cuddle
and
then
you
do
some
other
stuff
and
installation.
A
A
Ok
I:
do
you
think
it
would
be
I
think
we
need
to
resolve
this
sometime
between
now
for
the
0
3x
patch
stream,
or
in
0
for
I'm
gonna
put
it
in
0,
3,
X
4
right
now.
I
also
do
think
that
we
can
use
some
additional
thinking
and
design
on
this.
So
if
you
are
interested
in
this
topic,
I
would
encourage
you
to
take
a
look
at
Fabrizio's
suggestions
here
and
add
comments.
If
you
have
any
opinions.
A
Alrighty
next
up
is
e
to
e
jobs
for
cluster
cuddle.
I
think
we
should
do
this
as
soon
as
we
can
get
around
to
it.
So,
let's
put
it
in
0,
3
X.
Is
this
something
that
we
could
maybe
break
up
and
split
up
the
works
so
that
most
people
could
work
on
it
or
is
it?
Is
it
really
like
each
one
builds
on
the
next.
J
A
A
M
A
N
M
A
Thank
you,
everybody,
and
then
we
have
Daniel's
one
that
we
were
talking
about
before
I
think
we
need
to
put
this
one
either
in
next
or
zero.
Four,
because
we're
not
gonna
change
the
name
in
the
v-103
API,
so
I'll
put
it
in
zero,
four
and
I
believe
other
than
figuring
this
one
out
I'll
put
this
in
zero
3x
in
case
it
is
an
issue
and
then,
as
you
all
triage,
we
can
pump
it
out
or
close
it
or
do
whatever
is
needed
all
right.
That's
the
end
of
the
issues
that
need
triaging.
A
So
we
have
about
20
minutes,
left
I
will
repeat
my
PSA
from
I
guess
it
was
last
week
or
the
week
before
that.
If
you
do
have
any
features
that
you're
interested
in
seeing
beyond
0.3
slash,
it
went
out
three,
please
either
add
them
or
bring
them
up
with
the
next
meeting
or
chat
with
us
on
slack
that
we
definitely
want
to
start
planning
things
that
we
want
might
want
to
do
in
our
next
release.
So
next
my
release
so
now's
the
time
to
be
thinking
about
that
and
last
call
for
topics.
E
E
Probably
the
most
disjoint
demo
you've
ever
seen,
but
you
know
look
at
that
so
ivory
simulated,
where
we
were
before
we've
got
the
temp
sent
down
here.
The
notes
not
ready
and
I'll
show
you
the
difference
between
the
patch
commands
that
makes
this
work.
So
this
is
the
one
I
did
earlier
notice
on
the
left
there
just
under
where
it
says:
J
speed,
MacBook
Pro,
that
one
is
quoted.
If
you
take
that
away,
it
works.
E
So
that
is
something
I'm
going
to
look
into
and
probably
raise
a
bug
about,
because
I
don't
know
why
that
should
make
a
difference.
Thinking,
because
it's
that
int
or
a
percent
type
yeah.
That's
that's
gonna,
be
where
I'm
gonna
go.
Look
but
I,
don't
think
it
should
make
a
difference,
but
anyway,
if
I
do
it
without
the
quoting?
What
we'll
see
is
the
Machine
health
checkers
then,
like
our
good
I
can
now
go
ahead.
E
Deletes
the
machine
machine
set,
says:
okay
machines
deleting
brings
up
a
new
one,
which
is
a
second
one
in
this
list,
but
notice
gone
because
there's
no
workloads
on
it,
it
would
have
been
drained.
Had
there
been
any
and
eventually
the
cluster
will
settle
back
to
normality,
but
basically
what
I
had
left
from
that
demo,
but
I
got
stuck
by
that
protein
I'm.