►
From YouTube: Kubernetes WG LTS 20190205
Description
https://git.k8s.io/community/wg-lts
Meeting agenda / minutes: https://docs.google.com/document/d/1J2CJ-q9WlvCnIVkoEo9tAo19h08kOgUJAS3HxaSMsLA/edit?ts=5bda357d#bookmark=id.bzr3lyqt46x
A
A
A
A
All
right,
it's
two
paths
I'll
go.
It's
start
today
is
February
4th
2019.
This
is
working
group
LTS.
The
meeting
is
being
recorded,
we'll
upload
it
to
YouTube
the
meeting
minutes.
Agenda
topics
are
in
the
Google
Doc,
as
always
ask
that
everybody
adhere
to
the
community
code
of
conduct
and
be
good
people,
so
I
guess
first,
if
just
the
reminder
to
drop
your
name
in
the
attendees.
A
Second,
we
exist
two
weeks
ago
we
sort
of
talked
about
whether
we
keep
pushing
on
work
group,
formalization
or
not,
and
frankly,
by
last
week,
I
was
kind
of
done
with
putting
time
into
that
and
it
pretty
much
kind
of
decided.
I
wasn't
gonna,
be
giving
it
any
more
time
and
then
suddenly
things
came
together,
so
the
TR
merged
it
got
sufficient
lgt
Em's
from
people
who
wanted
to
see
at
least
the
discussion
move
forward,
see
where
things
go
so
hooray.
A
Next
thing
on
the
agenda:
the
survey
draft:
we
spent
a
lot
of
time
talking
about
this
in
the
last
meeting,
and
then
there
were
a
bunch
of
comments
on
the
doc
I
think
we've
mostly
resolved
them,
but
the
one
the
bigger
picture
thing
that
came
out
of
it
is
like.
Is
this
too
complicated
and
I
I?
Think
there
were
multiple
voices
saying
yes,
but
I
know
Josh
you'd
had
some
thoughts
on
that.
Do
you
want
to
share
them.
B
Well,
the
one
thing
was
that
I
kind
of
feel
that
this
surveys
a
candidate
for
having
two
different
decision
trees.
You
know
one
where,
if
somebody
says
they're
a
hacker
or
a
vendor
or
a
host,
you
know
somebody
working
for
kubernetes
company.
They
go
down
one
set
of
questions
and
that
set
of
questions
has
to
do
with
what
they
see
happening
in
their
general
user
base
and
the
other
one
is
if
somebody
is
an
end-user
or
consultant
or
if
someone
else
is
going
to
have
hands-on
production
called
clusters
that
they
go
down.
B
A
Makes
sense
to
me:
I
am
I,
don't
know
if
that
I'm,
like
I'm,
not
a
data
scientist,
so
I
don't
know
like
to
what
extent
that
could
change
what
we
learn
from
the
data.
But
if
we,
if
we
really
think
there's
two
disjoint
sets
of
people,
half
and
half
fish
of
the
the
questions
are
irrelevant,
then
it
points
that
way
well,.
B
Or,
to
put
it
another
way,
if
all
recipients
were
going
to
receive
all
questions,
I
would
actually
want
to
rephrase
some
of
the
questions,
because
I
mean
like,
for
example,
what
upgrade
philosophy
do
you
prefer
right
for
a
an
end
user?
That's
a
perfect
question
for
somebody
who
works
for
a
vendor.
That
question
really
needs
to
be.
B
What
upgrade
philosophy
do
your
customers
prefer,
you
know,
or
do
most
of
your
customers
prefer
right
because
that's
going
to
be
their
answer
and
otherwise
what
happens
is
when
you
have
a
question
that
doesn't
quite
target
the
person
who's
answering
it.
They
often
either
tend
not
to
answer
or
to
give
an
answer
which
is
basically
a
random
pick,
which
then
throws
off
your
results.
A
And
even
that
example,
I
feel
like
there's
more
nuance
there
that
could
be
informative,
like
the
the
customer
focus
is
definitely
good,
but
then
the
vendors
I
wonder
if
they've
already
sort
of
mentally
segmented
and
our
vending
towards
their
current
customer
set
because
of
the
way
they've
chosen
to
attack
it
from
an
engineering
perspective.
Well,.
A
A
So
it's
like,
then
it
becomes
two
questions
like
what
is
your
opinion
as
a
vendor,
but
then
what
do
you
see
reflected
in
your
cussed
or
bass
if
it's
different
it's
hard,
because
a
lot
of
these
are
kind
of
fishing
and
we
we
want
to
guide
the
answer,
a
certain
direction,
but
also
fish
open-ended
Li.
The.
B
Other
thing
that
I
would
like
to
do
because
I
don't
see,
I,
don't
see
getting
away
from
making
this
a
relatively
long
survey
right.
The
only
way
that
we
can
actually
make
it
a
short
survey
is
if
we
did
a
series
right,
if
we
actually
chopped
it
up
into
three
or
four
surveys,
that
we
were
going
to
do
over
a
few
months
which
we
could
do
the
drawback
to
that
is
that
you
know
the
same
group
of
people
want
answer
all
of
them,
and
so
we
kind
of
lose
the
correlation
factors.
B
I
I
think.
Instead,
a
better
thing
to
do
would
be
to
offer
some
kind
of
prizes
figure
out
some
kind
of
prizes.
We
can
get
funded
by
somebody,
the
CNCs
somebody
else
and
then
I
could
probably
put
up
one
prize
and
you
were
drawing
and
that
that
gives
people
that
won't
get
more
people
to
answer
the
survey.
But
what
it
will
do
is
it
gives
people
an
incentive
to
finish
the
survey
once
they've
started.
D
Discuss
in
the
past,
where
we
thought
of
using
this
as
the
first
survey
and
then
maybe
release
another
survey
later
and
what
we
learn
from
this
data,
so
it
was
like
this
is
the
first
time
we
are
doing
a
survey.
We
don't
have
a
lot
of
experience
for
doing
surveys
as
a
group
and
also
as
a
project,
so
maybe
slowed
this
out
sooner
and
then
learn
from
this
experience
and
build
a
better
serve
the
next
time
or
more
deeper
survey
that
cuts
the
cake
even
more
finer.
A
C
A
A
Of
that
is
visualization
like
the
multi
troit
multi,
how
you
do
stuff,
but
yeah
I.
Think
that
points
at
a
unanswered
decision
like
what
tool
are
we
gonna,
try
to
put
it
in
and
see
like?
How
long
does
it
turn
into
like
a
twenty
page
questionnaire,
and
do
we
estimate
it
takes
20
minutes
or
30
or
40,
or
what
to
go
through
da
ville'?
Had
you
played
around
with
putting
the
questions
in
a
tool
and.
B
Play
the
challenge
with
using
the
CNC
F
SurveyMonkey
instance.
Is
that,
for
whatever
reason,
only
C
and
C
F
full-time
staff
are
allowed
to
actually
have
their
hands
directly
ona,
which
is
going
to
make
it
difficult
to
do
a
complicated
branching
survey,
because
you'll
have
to
talk
a
CNC
F
staff
member
through
each
set,
rather
than
being
able
to
try
things
and
see
how
they
work
out.
D
E
D
E
A
So
I
guess
we
we
need
a.
It
would
be
good
if
we
made
at
least
kind
of
a
a
general
go
no-go
on
the
set
of
questions
ish.
But
then
so
we
know
that
there's
not
gonna
be
a
major
major
and
it's
still
to
come.
But
then
trying
I
guess
have
a
couple
of
us
put
it
in
Survey,
Monkey,
Google
Forms,
maybe
try
and
make
a
branching
and
see
what
it
feels
like,
because
that
that's
like
a
significant
time:
investment
for
somebody
to
go
mucking
around
with
those
specific
tools.
A
D
A
B
B
D
G
Yeah
I'm
not
sure
how
we
should
coordinate
the
thumbs
up
ODT
in
process
for
other
server
questions,
but
yeah
I
mean
I.
Do
think
that
it's
probably
it's
probably
past
time
that
we
get
into
some
sort
of
implementation,
I'm,
worried
about
the
perfect
being
and
env2
could
with
this,
because
it's
taken
a
long
time
you
know
to
get
so
they
sort
of
ready,
I
think
that
some
information
is
gonna,
be
better
than
you
know,
waiting
another
month
or
so
on.
Getting
it
exactly
perfect.
D
A
Okay,
well,
in
that
case,
double,
if
you
can
work
with
the
CNS,
you
have
people
to
see
and
maybe
even
try
on
a
free
account
just
to
compare
like
how
how
undermined
it
would
be.
I
guess
on
a
free
one,
just
to
compare
and
David.
If
you
could
try
putting
them
in
a
Google
form
it
and
then
send
a
message
to
the
the
mailing
list,
and
we
can
have
a
few.
Have
people
kind
of
look
at
how
it
feels.
A
And
then,
hopefully
we
if
we
can
iterate
on
that
asynchronously
over
the
next
week
or
so
than
in
our
meeting
in
two
weeks
time.
Hopefully
we
can
kind
of
feel
like
we
have
something
workable
and
maybe
best-case
B
thumbs
up
sending
it
out
in
two
weeks
time,
but
I
totally
hear
what
Nick
cells,
like
the
we've,
been
discussing
specific
questions
for
almost
eight
weeks
now.
So
at
some
point
we
gotta
move
it
forward.
Take
enough.
F
A
Even
just
like,
hey
I
know,
so-and-so
runs
a
cluster,
make
sure
that
they
have
filled
the
thing
out
and
just
kind
of
Bester
people
a
bit
to
where
we
we
can
to,
instead
of
just
sending
it
out
to
mailing
lists
and
hoping,
but
to
be
a
little
more
hands-on
on
people,
because
we're
not
looking
for
like
a
exactly
a
scientific,
random
distribution
of
responses
as
much
as
like
trying
to
get
actual
users.
So
if
you
know
somebody
getting
them
to
respond
is
valuable.
A
G
A
Okay,
well
sure
we
we
move
on
to
the
next
topic,
then
so
Jordan
I
saw
you
typing
some
stuff
there
and
I'd
gone
and
looked
at
the
issue
and
I
didn't
see
like
anybody
having
stepped
up
just
yet
on
ownership.
But
I
feel
like
that's
like
a
there's
like
a
project
management
ownership,
ownership,
II
part
here
like
to
delegate
and
and
see
if
we
can
have
some
specific
people
like
reaching
out
to
specific
SIG's
yeah.
C
Most
most
people
probably
don't
see
it
have
notifications,
and
that's
all
that
they
would
have
seen
from
these.
So
far,
so
I
think
there
are
I,
think
signaled,
sig,
ABI
machinery,
cluster
lifecycle,
release
and
testing,
maybe
scheduling
so
basically
the
the
ones
that
kind
of
own
the
release
and
cluster
lifecycle
process
and
then
the
ones
that
own
these
particular
components
being
discussed,
reaching
out,
at
least
in
slack,
maybe
even
in
one
of
those
meetings,
would
be
helpful.
C
So
if
there
are
people
who
are
already
attending
those
meetings,
you
want
to
kind
of
take
these
topics
the
relevant
ones
from
that
umbrella
issue.
That
would
that
would
be
helpful.
I
can
do
the
ones
for
cigar,
ch
and
say
gay
bi
machinery,
so
I
will
throw
my
name
in
here
and
if
people
want
to
grab
others,
node
scheduling
like
old
test
release,
putting
names
here
would
be
helpful.
A
Call
to
everybody
on
the
list:
if
you
glanced
through
the
set
of
the
six
issues,
six
seven
I
can't
count.
Seven
look!
If
you
tell
into
the
individual
ones,
you'll
see
the
label
for
sick,
whichever,
if
that's
a
cig,
with
which
you
have
an
affinity
to
ask
you
to
maybe
take
that
one
up
onto
the
edge
into
the
next
meeting.
It's
like
Lou
and
Mary
I
know
you're
sick
cluster
lifecycle
associated
the
one.
A
Of
if
the
notifications
are
missed,
it
gives
a
cookie
crumb
trail
for
the
inner
linkages
of
things
after
the
fact
to
think.
If
somebody's
looking
at
the
issue-
and
they
see
a
link
to
the
the
minutes
of
cig
foo
on
such
and
such
a
date
they'll,
they
might
be
able
to
go,
find
conversation
or
the
deserve
me.
Ting
is.
D
A
C
Yeah
I
think
the
the
documentation
thing
is
trying
to
go,
get
knowledge
out
of
people's
heads
and
get
them
to
write
it
down
and
make
sure
that
the
same
knowledge
and
expectations
match
and
where
there's
mismatches
we
surface
those
and
figure
them
out
so
I
think
the
very
act
of
writing
a
cap
would
do
many
of
the
same
things.
This
is
trying
to
do,
but
would
probably
be
overly
formal,
like
we're
not
actually
trying
to
change
anything,
we're
trying
to
understand
what
is
it
that
half
the
people
in
the
project
already
assumed
everyone
was
doing.
D
C
Yeah
I
don't
want
to
get
the
documentation
of
current
state
bogged
down
in
agreement
about
how
things
should
change
like
the
in
that
first
pass
was
very
useful
to
get
merged
into
the
website.
I
think
getting
these
details
described
and
nailed
down
as
well
as
quickly
as
possible,
would
be
very
useful
as
well.
E
A
Lots
of
comments
and
discussion
like
if
you
look
at
the
multi-tenancy
working
group,
they
had
maybe
four
or
five
different
documents
that
lived
over
many
many
months,
discussed
back
and
forth,
and
things
trying
to
bubble
up
and
common
eyes
out
of
them.
I
think
the
Google
Docs
as
much
as
it's
not
open
to
everybody
is
a
really
handy
platform
for
enabling
that
sort
of
collaborative
drafting
and
early
early
days
of
things,
I
think.
D
Because
I'm,
the
new
single
is
process
that
you
need
to
have
caps
for
each
of
those
issues.
Before
they
get
marked,
you
will
have
a
major
cap
from
working
group
and
then
each
of
this
github
issues
would
have
caps
caps
caps.
So
anyway,
all
the
owners
would
be
like
I
mean
there's
an
overhead
of
writing
a
lots
of
caps
for
this.
Just
so
I
think
I
agree.
Maybe
we
can
use
this
issue
and
then
start
fighting
motivations.
A
Think
will
yes
word:
they
immediately
assume
some
big
grand
vision
for
a
type
of
support
stream,
but
one
of
the
things
that
I've
liked
and
the
enhancements
process
right
now
in
114,
with
people
having
to
propose
caps,
there's
a
lot
of
smaller
caps.
It's
it's
more
point,
specific
improvements
that
are
easier
to
read
and
digest
and
then
get
approved
for
implementable
states
and
start
making
movement
towards
completion
so
that
that
might
be
a
model
to
keep
in
mind
trying
to
to
make
sure
that
we're
doing
smaller,
simpler
stuff.
A
D
D
A
Kind
of
related
to
this
I
was
reminded
yesterday.
People
were
kind
of
emailing
and
tweeting
in
slack
mentioning
the
CNCs
annual
annual
report
and
it
had
a
page
about
API,
snoop
and
I.
Wonder
if
that's
something
that
we
should
be
looking
at
here
as
well,
I,
don't
know
if
that
is
something
Jordan.
They
had
kind
of
informed
any
of
your
thinking
about
this.
G
D
A
One
of
the
things
that
specifically
caught
my
attention
was,
if
you
go
to
their
their
github
page,
there's
like
a
embedded,
Google,
spreadsheet
and
I
downloaded
a
copy
of
that,
and
then
just
like
did
a
grep
for
alpha
L,
a
CSV
copy
grep
for
lines
that
begin
with
alpha
beta
and
stable
and
piped
it.
To
word
count
a
shell
and
the
of
the
tracked
API
is
it's
it's
almost
half
in
half
beta
and
stable,
but
that
that
doesn't
us.
A
C
Is
the
plans
to
finish
flushing
out
the
stuff
in
the
extensions
B
1
beta,
1
API
group
and
like
get
a
plan
to
stop
serving
those,
probably
in
the
116
timeframe,
they've
been
deprecated
since
20
or
1/9,
and
so
hopefully,
as
we
start
kind
of
removing
some
of
those
long
deprecated
beta
versions
will
get
better
signal
there.
So
that
maybe
we'll
realize.
Oh,
you
know
what
actually
3/4
of
the
things
that
we're
verifying
are
stable
and
the
things
that
people
are
depending
on
they're
stable.
A
C
C
C
And
with
that,
in
particular
reaction
just
hit
a
milestone
where
we
have
green
CI
with
the
extensions
all
the
deprecated
workload,
API
is
disabled,
so
everything
that
is
required
for
CI
and
conformance
and
all
the
add-ons
and
add-on
manager,
and
so
so
many
things
required
updating
that
that
has
all
been
done.
So
we
have
green
CI
with
this
now,
which
is
awesome.
A
That's
cool,
so
one
thing
that
I've
kind
of
felt
unclear
on
it
seems
like
it's
been
a
steering
committee
level
focus
to
push
on
performance
like
there's
a
sub-project
formally
around
that,
but
it's
are
they're
like
how's
the
project
management
aspect
of
stuff
working
like
there's
a
couple
of
these
mega
issues
or
how?
How
could
people
get
involved
in
contributing
on
that?
All
of
that
work
or
know
that
it's
there
discover
it
next
tasks.
D
C
Yeah,
it
kind
of
depends
on
what
direction
you
come
at
it
from
if
you
are
coming
at
it
from
the
top,
and
you
just
say,
like
I'm
interested
in
conformance.
How
can
I
help
like
there's
lists
of
things
around,
like
some
of
the
Windows
stuff,
trying
to
figure
out
how
to
navigate
compartments
issues
across
platforms?
There
is
some
of
the
beta
GA
questions
about,
like,
as
things
go
to
GA,
making
sure
they
have
the
representative
formants
if
they're
required
and
then
the
question
of
what
do
you
do
for
things
that
are
GA
but
optional?
C
So
there
are
some
kind
of
high
level
questions
there
if
you're
interested,
if
you're
coming
at
it
from
the
bottom
right,
you're
you're
working
on
a
particular
feature
and
you're
saying:
oh,
let's,
let's
graduate
this
kind
of
the
same
questions
but
from
the
opposite
direction.
I
guess
is
this
a
thing
that
is
expected
to
be
in
all
clusters?
If
so,
then
it
needs
a
conformance
test
to
represent
it.
Is
it
optional
than
it
needs
ete
tests
and
it
kind
of
goes
in
that
bucket
of
like
we
have
e
to
e?
D
C
So
there's
a
the
deprecation
policy
goes
into
a
fair
amount
of
detail.
I
will
link
that
as
well.
Ap
eyes
are
the
best
defined
because
they
have
a
lifecycle
and,
like
a
time,
timeline
associated
with
it.
There's
kind
of
a
catch-all
bucket
at
the
end
of
that
document
that
says
bait
behavior
that
is
not
associated
with
an
API
or
a
command
line.
Flag
should
function
for
a
year
after
announced
deprecation.
C
There
still
needs
to
be
a
little
bit
more
guidance
around
like
how
do
you?
How
do
you
get
feedback
on
why
it's
being
removed?
Is
it?
Is
it
being
removed
because
it's
causing
and
technical
problems,
if
so,
then
to
communicate
it
out?
Well
it
being
removed,
because
we
think
no
one's
using
it?
If
so,
how
do
we
know
that
no
one's
using
it
did
we
actually
check
so
that
there
could
probably
be
some
better,
better
guidance
there
about
due
diligence
as
we're
kind
of
waiting
out
that
year,
so.
D
D
D
D
D
F
A
A
I,
don't
remember
everybody's
affiliations,
but
as
we
were,
one
of
the
things
that
came
up
in
merging
the
into
the
the
cigs
animal
was
we
needed
a
a
quorum
of
steering
committee
votes
in
favor
and
looking
through
the
steering
committee
list,
it
dawn
on
me
that
they're
sort
of
looking
at
the
spread
of
companies
and
vendors
that
are
represented
on
the
steering
committee
list.
Those
larger
vendors
aren't
all
represented
so
far
in
activities
in
this
workgroup.
So
Microsoft
is
one
that
stood
out
in
particular
where
we
hadn't
gotten
people.
A
D
A
A
D
A
A
D
And
I've
seen
that
sick
chairs
are
not
always
represented
up
there,
entire
six
on
all
topics
right,
so
it's
good
to
go
and
talk
to
some
of
the
SIG's
at
least
the
major
vendor
is
at
least
when
those
who
have
a
sick
cloud
provider.
There
are
in
three
or
four,
so
it's
it's
easier
to
go
there
and
talk
to
everyone
and
have
a
wider
audience,
especially
when
we
have
the
survey
it's
much
easier
out.
There.
A
A
So
Donald
and
David,
if
you
do
sort
of
a
draft
and
send
it
out
sometime
over
the
next
week,
just
to
the
the
working
group
mailing
list,
we
can
provide
feedback
asynchronously
and
see
what
we
can
get
drafted
up
for
for
the
next
meeting.
It
was
me
all
right
anything
else.
Anybody
wants
to
talk
about
last
call:
okay,
cool
thanks,
I'll
upload,
the
video
and
a
bit
and
good
rest,
your
name
and.