►
From YouTube: 20180725 sig cluster lifecycle kubeadm 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).
A
A
B
B
A
A
Regarding
some
of
these
things,
no
one
owns
the
controller
manager,
so
it
would
be
like
we
would
have
to
do
that
yeah.
So
so,
I
don't
exactly
know
how
we
proceed
other
than
like
someone
from
the
sig
literally
going
through
and
helping
to
move
some
of
the
configuration
changes
unless
Stefan
is
committing
to
making
that
effort.
Go
I.
A
B
B
A
Here's
a
question
for
you:
do
you
think,
like
cig
coaster,
life
cycle
should
Co
maintain
some
of
that
stuff,
because
it
moves
it
out
of
the
hands
into
more
broader
configuration
for
the
components
and
if
they
change
things
at
least
being
aware
of
those
changes,
makes
a
lot
of
sense.
The
problem
is
like
it's.
An
API
it'll,
be
an
API
Machinery
staging
area
unless
you're
totally
fragmented
into
the
a
completely
separate
repo.
So.
B
B
A
Also,
like
the
broader
problem
with,
like
general
configuration
of
distilling
it
down,
it's
like
there's,
common
configuration
exhibit
like
I
was
just
talking
with
Clayton
about
this
the
other
day
and
that
there
are
a
set
of
non-overlapping
concerns
between
features
and
common
things
that
should
occur
across
every
component.
Right
like
there
should
be
a
shared
set
of
component
configurations
for
every
component.
Those
include
things
like
GPS,
you
know
and
other
parameters
that
exist
across
all
demons,
but
they're
not
right.
Now,
they're
kind
of
like
spurious
lis
located
in
all
these
different
locations.
B
Yeah
exactly-
and
that
is
what
this
separation
of
code
or
code
refactor
is
hoping
to
soul
to
make
it.
So
we
have
that
this
area
for
channel
now
we're
gonna
put
something
in
case
at
io,
/,
API
machinery,
we're
gonna,
add
one
new
config
group
with
you
till
these
digital
strikes
like
okay.
Here
is
how
you
connect
to
the
IP
eyesore
and
then
in
case
a
tile
flash
API.
So
we
have
one.
B
B
Let
any
component
like
the
cubelet,
Control
Manager
schedule
also
needs
mm-hmm,
and
then
we
have
some
packages
that
move
out
the
case
at
I/o
utils,
but
there
are
just
like
code,
that
is,
that
is
code,
but
I
noticed
something
Tim
this
more
today,
this
case
at
I/o
util,
doesn't
even
have
all
this
and
on
this
file.
So
it's
yeah,
it's
an
a
blurry
state
but
we're
hoping
to
eventually
figure
things
out.
I'm
gonna.
A
B
B
B
A
A
But
I
think
for
the
long-term
maintenance
and
support
ability
for
configuration
of
these
components.
I'm
surprised
that
Google
went
heavily
invest
in
some
of
these
things.
I
got
to
figure
out
how
to
social
food
that
one,
because
it
makes
a
ton
of
sense
that
people
want
to
be
able
to
customize
configuration
for
deployments
in
any
way
shape
or
form,
especially
for
the
cluster
API
perspective
right
yeah.
B
B
A
B
B
Let
me
say
this
way
when
we
have
the
initial
API
group
there,
myself
and
Stephane,
and
some
other
involved
in
this
in
those
files
we
can
hack
around
and
build
stuff
and
when
we're
happy
with
the
like
structure
and
everything
we
can
send
the
next
cap,
which
is
okay.
Now
this
is
now
we've
learnt
X,
Y
and
C,
and
we
are
ready
to
move
to
data.
And
now
we
go
through
the
full
API
process
for
every
every
con
things.
B
A
C
A
What
Liz,
myself
and
for
bricio
walkthrough
at
all
the
old
I'm,
not
concerned
I,
like
the
approach
I,
think
that
the
biggest
concern
I
have
is
more
from
a
consumer
/
UX
perspective
than
it
is
from
the
plan.
The
plan
I
think,
makes
a
lot
a
ton
of
sense
from
a
sort
of
software
design,
perspective,
I,
think
from
the
UX
perspective,
what
users
expect
and
what
they're
used
to
it
is
a
pretty
large
departure,
so
I
do
think.
A
So
we
need
to
make
sure
that,
like
when
we,
when
we
have
release,
notes
for
the
configuration
changes
that
we
spell
out,
that
this
is
a
migration
and
inherently
migrations
can
be
dangerous
or
non
compatible
in
some
ways
and
that
they
need
to
make
sure
they
vet
their
configuration
before
they
do
the
upgrade
itself
to
make
sure
it's
all
seen.
Yeah.
B
A
D
Think
the
I
think
that
mostly
just
echoing
what
Tim
said,
our
biggest
our
biggest
problem
is
Docs.
We're
going
to
need
to
make
sure
we
write
a
lot
of
documents
just
like
this
is
how
the
new
thing
works.
This
is
like
what,
if
you
used
this
setting
before
this,
is
where
you
would
find
it
now
all
that
sort
of
stuff.
So
I
don't
want
us
to
think
once
we
get
the
the
the
PRS
done.
B
A
B
B
D
D
A
The
change
for
the
configuration
is
like
a
long-term
maintenance
support
for
us,
so
it's
not
really
like
a
user
driven
change.
It's
like
how
can
we
design
the
configuration
such
that
we
can
support
this
in
the
long
haul
and
not
suffer
all
the
pits
and
follies
of
that
we've
encountered
I'm
because,
as
we
once,
we
get
to
beta
we're
gonna
be
stuck
with
this
for
a
long
time.
A
D
I
mean
that's,
that's
just
like,
even
if
the
change
itself
is
is
not
user
driven,
we
don't
want
to
finalize
the
beta
can
fake,
and
once
we
get
there,
a
bunch
of
users
are
just
like
this
doesn't
work
at
all
for
our
use
case.
This
is
not
helpful
for
us
that
sort
of
thing
I
think
it
would
be
really
good.
If
we
could
to
find
somebody
to
look
this
over
and
say,
yeah
I
would
be
able
to
use
this.
A
A
There's
no
there's
no
real
students
for
how
we
go
about
doing
this
in
a
clean
manner
that
doesn't
chaotic
I
can
imagine,
unleash,
and
chaos
be
like
hey,
here's,
a
camp
and
Twitter
tweet
it
out,
and
then
you
get
like
the
randomness
of
the
internet,
which
is
not
productive
either.
You
know
what
SWAT
like
a
set
of
trusted.
A
D
B
E
B
G
G
D
So
probably
not
specifically
useful
for
getting
feedback
on
this
proposal,
but
uncle
I
agree
in
general.
I
always
try
to
push
back
against
engineers
who
build
stuff
for
other
engineers,
making
UX
decisions
because
we
have
a.
We
have
a
tendency
to
put
blinders
on
and
be
like.
Well,
this
is
going
to
make
the
codis.
A
Hit
the
nail
on
the
head:
that's
kind
of
what
my
biggest
concern
was.
What
some
of
the
changes
were,
because
when
people
depend
upon
the
current
UX
of
a
tool,
we
need
to
make
sure
that
the
consistency
is
maintained
for
that
UX
and
and
we're
listening
to
their
needs.
The
problem
is
like
the
randomness
of
the
internet
with
regards
to
this
particular
project.
A
D
D
D
E
E
B
D
D
B
B
What
specific
structure
we
put
in
the
joint
configuration,
for
example,
is
to
be
discussed,
but
I
think
the
in
its
configuration
is
it's
already
kind
of
approved
from
our
cycle,
with
no
registration
options
and
bridge
of
circles,
so
that
is
kind
of
done.
The
other
two
needs
needs
more
polishing
to
be
decided
in
the
cap,
but
if
we
at
least
make
offer
tree
the
version
that
splits
out
this
stuff
and
then
we
can
iterate
on
on
the
actual
I,
don't
know
field,
grouping
and
stuff
inside
of
classic
version.
A
All
right
next
item
is:
it
relates
to
your
topic,
but
we
we
have
technical
debt
with
bootstrap
tokens
that
we
were
supposed
to
face
and
collect
and
client
go
as
well
as
I'd
like
to
rip
it
out
of
our
internal
structures
because
it
doesn't
belong
there.
We
already
knew
that
so
I'm
gonna
bump
the
issue
to
p0
and
probably
work
with
Stefan
to
try
and
find
a
happy
landing
home
for
this
stuff,
but
ideally
I'd
like
to
rip
it
out
of
tube
ATM
types
because
it
doesn't
belong
there.
Yes,.
B
A
A
B
I
mean
what
would
exist.
There
is
the
types
that
represent
the
secrets,
then
so,
like
muta
token,
distracts
all
the
functions
that
relate
to
passing
a
bunch
of
token
and
converting
it
into
a
secret
and
do
run
trips
back
and
forth,
and
maybe
this
doesn't
exist
and
we
maybe
want
to
create
a
client
ish
thing
for
it
as
well.
I
don't
know
if
we
need
that
actually
but
yeah.
So
those
functions
that
I
don't
know
what
other
functional
we
could
break
out
our
core
into
butchart.
B
We
may
want
to
break
out
the
booty
up
token
stuff
in
the
controller
manager,
the
actual
like
just
a
package.
You
know
and
then
still
run
it
vendor
it
from
the
control
manager
code,
but
bender
it's
from
the
bishop,
repo
and
then
Sigma
slicer
would
be
the
kind
of
canonical
owner
of
this
repo
I
think
that
was
what
most
people
like
sin
on
the
thread.
A
B
Yeah
and
it's
it's
mostly
a
pure
code
move,
so
so
it's
not
that
tricky,
but
it's
still
if
there
are
a
lot
of
small
communications
indeed,
but
this
is
in
the
end,
this
is
gonna.
Look
like
or
look
very
similar
to
the
common
config
struct.
We
do
we
split
out
in
the
proposal
we
first
talked
about
here
which
regards
the
component
configs.
It
is.
It
is
pretty
much
gonna.
Look
like
that.
So.
B
A
Just
had
a
quick
pull
before
I
think.
The
next
topic
is
a
quick
poll
before
I
proceed,
I'd
like
to
rip
out
self-hosting,
just
like
all
the
code
and
maintenance
of
it
before
we
start
to
get
to
the
AJ
configuration
stuff.
It
just
simplifies
the
deployment
and,
if
we
want
to,
we
can
always
create
a
separate
script
or
utility
that
basically
does
the
pivoting.
If
people
desire
that,
but
once
they
do
that
then
you're
out
of
the
upgrade
life
cycle,
maintenance
of
rubidium.
A
H
B
A
B
A
We
need
more
resources
right
now.
We
are
like
hefty
of
shifting
back
to
it,
but
if
we
can
get
people
to
commit
to
stop
zombie
super-dee-duper,
the
burn
rate
is
not
so
great,
but
we
do
have
everything
triaged
so
like
one
one
twelve
cycle
as
all
the
issues
assigned
and
willowmere
is
doing
a
bang-up
job
with
like
addressing
inbound
as
soon
as
it
comes
in
and
I
think
we
just
need
folks
who
can
commit
to
doing
pieces
of
the
backlog,
not
just
the
pieces
that
they'd
like
to
do.
B
A
Dogs,
it's
not
even
a
scent
no
file,
it's
like
I
think
we
used
the
wrong
term
or
somebody
triggered
off
the
wrong
term.
It's
it's
like
it.
Just
transaction
history
right
that
I
they
took
a
issue.
They
ran
with
it
without
discussing
possible
solutions.
The
conversation
piece
we
had
was
that
it'd
be
nice
to
have
some
history
when
people
long
issues
against
qadian
rebuttal
of
all
the
details
of
what
they
had
run
their
argument
list
to
to
submission
so
like
we
could
put
have
as
part
of
the
major
filing
a
bug
request.
B
A
If
it's
localized
to
the
Machine
and
reads
var
Lib
Canadian
that
it's
done
for
a
lot
of
things,
it's
not
unheard
of!
So
that
way,
you
know
what
was
called
and
the
person
doesn't
need
to
add
it
to
the
submission
it's
up
to
them
if
they
want
to
or
not,
but
also
that
for
an
operator's
perspective.
If
you
had
multiple
operators
coming
through
a
given
infrastructure,
you
would
know
what
was
called
on
the
different
nodes
as
that
history
can
sometimes
be
lost
across
reboot
cycles
and
whatnot.
D
H
H
H
G
B
Okay,
oh.
B
A
Saw
it
today
so
I
need
to
take
a
look
at
it,
but
I
was
I
generally
thought
we
would.
We
wanted
to
shed
phases
and
move
it
into
the
sub
commands
for
cleanliness
and
from
I
haven't
even
looked
at
the
PR
yet,
but
the
the
name
of
the
PR
was
kept
for
cuvee
and
phases
to
beta,
which
I,
which
is
different
than
his
original
proposal
and
I.
B
H
A
B
A
A
B
Mm-Hmm,
okay,
I
yeah
I
could
actually
share
my
screen
here
and
just
show
the
planning
doc
that
we
had
open
and
see
where
we're
at
while
we're
talking
about
this.
So
with
regards
to
conflict
to
be
the
one,
this
is
kind
of
done.
We
need
some
more
editing,
changes
and
minor
stuff
here,
as
well
as
well
as
get
solid
feedback
from
the
rest
of
the
community,
the
broader
ecosystem.
B
A
B
B
A
Need
to
resync
up
with
jennifer
some
of
the
changes
have
already
been
made,
but
I
think
I
need
to
go
circle
back
with
Jennifer
to
figure
out
where
exactly
we
are
now
I
know.
I've
been
reviewing
a
number
of
small
pr's
in
some
critical
ones
to
the
docs
changes,
but
the
the
overall
grand
unified
field.
Theory
change
for
docs
I,
don't
know
what
state
that's
in
yet
I'll
poke
around
slack.
Okay,.
B
B
A
B
B
Yeah,
it's
it's
not
done
in
a
good
way,
just
like
that's
very
easy
to
say
right
now,
if
it's
like
three
months
to
exploration
time
or
something
and
you
upgrade,
it's
gonna
rewrite
the
search
after
the
upgrade
process,
but
this
would
be
something
more
like
and
that's
only
for
the
API
server
serving
cert,
but
this
was
this
would
be
more
like
okay,
executors
see,
and
it's
using
right
now
is
using
the
local
C
on
disk.
This
would
be
more
before
the
upgrade
starts,
execute
a
couple
of
CSRs
and
approver.
So
we
have
root
access.
B
We
can
approve
them,
so
they
assigned
then
right
on
to
disk
store
the
old
ones
in
the
same
file,
shifting
manner
we
have
and
in
a
backup
place
and
when
the
new
API
server
starts
up
the
new
version,
it
would
already
have
this
new
search
because,
right
now
we
the
idea,
so
it
still
uses
old
search.
It's
only
until
the
next
upgrade
that
it's
effective,
so
the
less
stuff
to
be
improved
there.
B
This
is
on
kinda.
Is
it's
moving
slowly
this,
but
it's
kind
of
on
track
with
the
proposal
that
I
have
merged
cuz,
yeah,
myself
and
Stephane
is
working
on
it.
He
will
send
a
PR
for
the
scheduler,
which
is
a
like
the
initial
initial
components
to
be
done.
Then,
after
that
everything
else
will
follow.
We
had
some
Chinese
contributor,
I
think
signing
up
for
moving
the
new
proxy
stuff
staging
as
well,
which
is
then
a
prerequisite
for
getting
with
the
beta
so
but
yeah.
B
B
A
Looked
that
so
we
had
people
sign
up
for
this,
but
then
they
would
prove
so
Lee
Lee
was
planning
on
helping
out
here
from
it
set
up
some
of
the
tests
and
one
of
our
internal
testing
resources,
who
was
actually
very
active
Ruben,
has
been
priority,
interrupted
towards
other
things.
I'm
I'm
poking
him
right
now
with
regards
to
like
when
he
can
come
back,
because,
ideally
we,
if
he
can
switch
back,
we
can
start
to
make
progress
of
some
of
the
test,
automation
that
we
we
already
have
the
jiggery
for
the
test
automation.
A
A
B
I
also
think
that
Ben,
the
other
said
something
along
the
lines
in
suggesting
or
sick
release
or
whatever
that
he
can
help
with.
With
this
kind
of
setup,
I
mean
he
has
very
extensive
background
in
suggesting.
So
really
really
knows
what
he's
doing
there
and
I
I
could
talk
to
him,
but
I
I
hope
we
may
be
get
some.
We
may
get
some
help
from
there
just
as
a.
A
Broader
testing
issue
thing
Google
is
finally
looking
at
building
out
sort
of
a
/,
PR
dinned
type
of
solution.
So
that
way
you
can
actually
executing
the
binaries
that
you're
that
you've
made
changes
to
and
I
am
trying
to
converse
more
broadly
with
fada
and
then
to
try
and
channel
some
of
this
work
so
that
we
could
get
a
use
cube
idiom
to
be
the
deployer,
the
default
deployer
for
that
jig,
and
that
would
be
a
Napper
PR
basis.
Then
yes,.