►
From YouTube: Kubernetes SIG CLI 20170614
Description
Bi-weekly meeting for the SIG-CLI Kubernetes group.
A
One
thing
that
we
see
is
there's
a
lot
of
contributions
from
the
community
or
other
states
to
coop
control,
but
they're
not
always
focused
on
kind
of
the
things
that
are
priorities
for
us,
so
they
might
be
other
commands,
for
instance,
I
think,
there's
debug,
commandments
being
added,
and
one
about
service
accounts,
number
of
other
things
and
we're
trying
to.
We
have
to
review
those
review.
The
proposals
try
and
figure
out
if
the
flags
make
sense
a
lot
of
times.
A
If
you,
if
you
try
and
write
the
documentation,
bradda
imperative
lis
create
things,
it's
not
fun,
because
we
have
to
basically
say
we
like,
haven't
really
completed
this
stuff,
and
we
don't
know
if
we're
going
to
complete
it.
Just
a
couple
of
things
they
are
like
to
control
run,
include
control,
create
deployment,
don't
have
the
same
set
of
flags,
even
though
they're
supposed
to
do
the
exact
same
thing,
and
so
we
can't
say
the
same
and
there's
we
don't
have
create
commands
for
all
the
stuff.
We
just
have
a
subset
of
the
workload
types.
A
B
And
I'd
love
to
see
us
use
the
create
stuff
as
a
way
to
sort
of
initialize
a
amo
file
on
disk
right,
so
it's
like
I
can
create
output
it
to
disk,
tweak
it
and
then
and
then
instantiate
it
right
to
start
to
try
and
draw
the
connection
between
those.
But
it's
easy
to
put
feature
requests
out
there.
That's
a
problem
see
that
now
and.
A
A
And
and
kind
of
on
the
same
note,
I
think
one
thing
we
probably
want
to
think
about
is:
what
commands
should
we
be
adding
at
this
point,
you
have
a
pretty
good
understanding
of
ok.
Could
control
should
manage
your
config
and
be
able
to
manage
objects
from
the
cluster,
so
the
imperative
great
commands
on
sacraments
and
view
commands
make
sense.
A
A
C
A
So
maybe
I'm
not
sure
if
there's
anything,
we
have
any
action
items
out
of
this
directly.
Here,
I'd
like
to
see
probably
a
document
around
what
kind
of
PRS
that
we
want
to
accept
what
kind
of
PRS
we
want
to
hold
off
on
until
we
have
a
better
story
around
kind
of
our
technical
debt
issues.
A
The
other
thing
I
was
thinking,
oh
go
on,
okay,
sorry,
the
other
thing
I
was
thinking
we
might
do,
and
then
this
can
come
into
play
in
the
repo.
So
on
the
repo
split
technical
that
we
have
to
complete
the
split
work
we're
gonna
have
to
do
is
like
trading
kind
of
PRS
like
if
you
want
to
see
a
PR
to
add
a
flag
to
particular
command
to
some
feature
like
help
out
breaking
one
of
the
dependencies
and
just
have
a
list.
I
have
a
LinkedIn,
a
wiki
page
about
200
dependencies.
B
So
the
whole
idea
of
this
yeah,
what
is
in
the
release?
What
is
out
of
the
release?
I,
don't
think
across
kubernetes,
were
raving
about
this
I
think
one
again,
I,
don't
know
if
you
can
track
in
the
sale.
We
hit
this
with
cube
admin
where
we
had
some
1.7
block
and
cube
admins
bugs
in
the
cube
admin,
refocuses
property
issues
worst,
those
were
not
being
tracked
by
the
release
team
they're
like
oh
look,
there's
no
bugs
that
are
blocking
there.
They
totally
have
looking
outside
of
the
main
and
I
know
now.
B
That
is
that
people
have
bugs
that
are
broken
out,
also,
maybe
a
similar
thing,
but
then
there's
this
question:
okay,
what
reclose
actually
need
to
be
tracked
for
part
of
the
release
versus
which
repos
are
on
a
different
release
in
April
and
that's
just
going
to
get
Messier.
Maybe
so
I
don't
have
any
answer
there
I'm
just
saying
like
it's
something
that
you
might
want
to
like.
We
need
to
pay
attention
to
and
we
probably
form
an
opinion
about
what
we
think.
The
right
thing
is
I'm
with
the
right
thing.
A
I
think
the
kind
of
the
action
I've
been
trying
to
push
things
is
ownership
of
all
the
release,
blocking
items
being
pushed
out
from
centrally
managed
Ingle
like
release
team,
because
it
does
lend
itself
towards
stuff
getting
drops
on
the
floor
like
every
release.
The
release
team
is
kind
of
guessing
at
what
tests
are
blocking
and
what
tests
are
not
blocking.
You
know
that
one's
been
failing
forever,
don't
worry
about
it.
It's
really
hard
to
keep
that
at
this
rate,
and
so
I
think.
A
A
A
A
A
There
we
go
so
there's
a
summary
page
and
you
can
kind
of
look
through
here
and
I.
Come
daily,
I
probably
check
this
and
look
through
and
say:
okay,
do
we
look
good
or
what
open
and
we're
trying
to
get
the
one
set
up
for
each
sake
or
each
area
of
ownership
at
least
right,
so
sig
Network
signal
to
storage
and
then
that
will
give
folks
more
control
over
the
stuff
they
own.
C
C
B
C
Exactly
and
also
you
know,
we
mentioned
as
the
first
topic
of
the
meeting
agenda
about
how
we
are
in
the
need
of
help,
for
you
know
reviewing
few
PRS
and
things
like
that.
I'm
reviewing
proposals,
it's
said
there,
so
something
I
didn't
know
yet,
but
I
think
it's
worth
doing
enough.
I'm
really
not
sure
how
much
of
a
result
that
would
have,
but
anyway,
I'll
probably
do
that,
send
an
email
to
the
public
lists,
probably
users
and
depth
and
calling
for
anyone.
C
You
know
interested
in
a
contributing
with
six
CLI
sub
CTL,
specifically
so
here's
here's
our
contributing
doc.
So
I,
don't
think
that
we
ever
made
that
public
of
you
know
public
knowledge
on
at
and
at
least
or
anything
like
that.
So
I'll
make
sure
I
do
that
and
hopefully
we
can
get.
You
know
some
new
people
to
help
a
little
bit.
B
Yeah
I
think
finding
ways
to
bring
more
folks
in
is
good.
I
was
just
commenting
to
Kelsey
here
that
it's
sad
that,
like
cute
control,
is
the
front
door
for
so
many
people
and
they're.
So
if
you
folks
working
it's
kind
of
a
tragedy,
so
I
wish
I
could
do
more
personally,
but
oh
well,
evangelize
at
least
yeah.
C
And-
and
actually
it's
very
interesting
because
you
know
for
really-
newcomers
is
a
not
too
hard
place
to
get
started
right,
so
helping
with
command
is
definitely
probably
much
easier
in
comparison,
so
starting
with
Adonal
controllers,
or
something
like
that,
so
it's
a
very
good
place
for
for
newcomers.
So
in
that
sense,
so
helping
with
basic
commands
fixing,
you
know
seem
simple
bugs
and
things
like
that.
It's
a
it's
a
very
good
starting
point.
So,
with.
B
D
B
D
C
Absolutely,
and
just
so
to
make
it
clear,
that's
a
direct
reflect
of
us,
you
know
be
short
in
terms
of
contributors,
because
I
was
speaking
to
to
Philip
earlier
this
week
or
last
week
that
we
are
really
working
a
lot
of
on
a
code
reviews
and
a
lot
of
on
purple
level
reviews,
but
we're
are
really
just
getting
much
more
than
we
as
a
team
are
able
to
handle.
So
it's
you
know
it's
a
it's
a
problem,
and
it's
not
nice
for
us
and
I
bet
it's
much
worse.
For
you
know,
contributors.
A
Yeah
I
think
this
is
like
a
really
good
example
of
I
was
talking
about
earlier
we're
into
the
arm
bug
obviously
of
the
night
it
in
today.
It's
not
on
the
most
urgent
list
and
probably
for
us,
or
what
I've
had
the
most
folks
wanting
to
get
resolved
so
trying
to
balance
those
priorities
is
challenging,
and
this
is
where
you
just
taking
this
bug
up,
because
you
had
a
specific
interest
in
fixing
this
issue
and
one
of
the
fixed
or
you
are
more
work
to
pick
up.
Yeah.
D
We
figured
this
out
because
we
wanted
to
forward
one
of
the
boards
from
the
Gloucester
owns
a
local
machine.
So
what
you're
developing
is
a
kind
of
a
development
tool
for
other
mobile
or
web
app
developers
to
visit
they're
back
at
all.
So
what
happened
was
we
need
to
have
a
way
of
people
to
you
open?
It
is
locally,
so
we
are
using
it
and
in
order
to
get
one
of
the
ports
of
mini
tube
forwarded
to
the
local
system
so
that
they
can
use
something
in
the
browser
we
needed.
D
The
signature
so
I'm
not
sure
if
I'm
making
sense.
So,
basically
what
happened
was
we
wanted
to
work
for
forwarding
to
stay
alive,
even
mentor
for
research,
because
our
architecture
is
in
such
a
way
that
any
configuration
change
restarts
the
poll,
so
the
tunnel
is
broken
as
we
can.
The
board
restart?
A
I
probably
need
to
look
do
some
more
reading
to
have
enough
context
before
I
understand
that
so.
B
I
I
do
think
that
this
is.
This
is
something
that
bit
me.
I
know
when,
when
I
was
working
on
the
the
book
that
Kelsey
and
I
are
are
working
on,
you
do
like
a
few
control
and
be
like.
Ok
now,
I
want
to
connect
to
it,
and
you
have
to
go
through
this
step
where
it's
like.
C
B
So
it's
like
there's
no
gap
there
and
you
end
up
having
to
actually
expose
people
the
details
that
they
probably
don't
care
about
to
actually
get
to
that
point.
So
I
think
from
usability
point
of
view.
This
is
this
is
a
little
bit
of
a
rough
edge
and
it's
definitely
I
think
from
a
from
a
capability
point
of
view.
If
you're
restarting
stuff,
you
don't
want
to
restart
it.
I
thing
here
is
from
a
depending
on
how
this
ends
up
getting
done.
B
It
could
be
something
that
hits
a
bunch
of
different
areas
right,
so
it
may
not
just
be
cute
control.
It
may
hit
some
other
parts
of
the
stack
server
side
to
be
able
to
really
support
this
and
so
I
think
it's
probably
a
little
bit
of
a
trickier
thing
to
design
and
they
make
work.
Then
then
it
might
first
appear.
D
B
That
proxy
stuff
is
really
a
nht
proxy
I,
sorted
looking
into
it,
based
on
your
bus
mate
on
your
email
and
actually
there
is
no
great
description
out
there
around
the
difference
between
to
control
proxy
versus
to
control
port
forward,
and
so
it
was
something
that
I
was
going
to
do
a
little
bit
of
research
and
try
it
right
out.
But
then
I
ran
out
of
time,
but
it's
not
obvious
in
terms
of
which,
which
one
you
use
in
which
situation
I
think.
Probably
the
mobile.
D
D
A
B
D
A
Okay,
great
I'll,
give
it
quick
just
update
on
the
repos
but
kind
of
the
work
that's
being
done
there
well
before
we
even
get
to
the
problems
that
kuba
admin
would
face
and
most
other
folks
other
faces.
A
Others
are
a
little
more
challenging,
such
as
utilities.
That
may
be
a
couple
different
folks
use,
but
our
relatively
self-contained
don't
depend
on
like
types
or
API
machine
or
anything,
and
so
those
could
be
potentially
moved
to
their
own
standard
library.
Repo.
That
Tim
is
talking
about
a
number
of
times,
there's
a
fare
list
of
those
things.
So
it's
not
a
not
not
going
to.
It
requires
a
little
bit
of
work
but
straightforward
at
least,
and
then
there's
other
dependencies
that
are
CalPERS.
A
Four
types,
so
these
depend
on
maybe
pods,
but
tell
whether
a
pod
is
healthy
or
not.
Right,
like
for
rolling
out,
is
just
quite
healthy
enough
to
roll
out
more
pods
or
something
like
that,
and
then
there's
this
whole
thing
about
inversion
types.
So
anyway,
there's
a
bunch
of
different
types
of
Simmons's.
We
have.
We
need
to
go
through
and
figure
out
a
bucket,
each
of
them
and
then
kind
of
come
up
with
a
story
of
how
to
solve
each
type.
So
so.
B
B
My
understanding
is
that
we're
starting
to
go
down
that
path,
but
I
think
four,
one
seven
is
optional
right
if
we
fully
complete
that
process
so
that
we're
you
know
driven
by
the
opening,
API
stuff
and
we're
you
know,
as
the
only
way
to
actually
do
stuff.
Does
that
simplify
the
set
of
dependencies
that
keep
control
depends
on
my
scientific
wildest
guess
is
20%
okay,.
A
C
A
B
And
then
like
I,
imagine
that
that
might
be
some
of
the
unburnt
types
type
of
stuff
right,
like
the
inversions
like
like
like,
because
you
know
cutting
the
cutting
of
the
dependency
on
the
inversion
types
of
the
healthy
thing
for
to
control
the
date
right.
Yeah.
A
In
theory,
I've
gotten
mixed
messages
from
API
machinery.
Child
says
you
have
to
have
the
Unversed
in
types
Daniel
says:
don't
use
them
the
reason.
The
reason
so
the
inversion
types
come
into
play
in
kind
of
two
places.
One
is
I
think,
probably
what
you're
thinking
about,
which
is
libraries
that
could
like
kubernetes
libraries
that
we
pass
in
the
types
yeah.
C
A
Who
want
to
break
another
way
they
come
into
play
is
I.
Think
Chow
was
saying
under
some
types
are
useful
for
if
you
don't
know
what
versions
of
a
object
server
supports
so
say
you
want
to
like
get
get
deployment
right
and
the
new
server
has
a
new
version
of
deployment.
If
you
use
the
version
versus
the
extensions
version,
a
if
you
use
the
unburdened,
it'll
fall
back.
If
you
talk
to
an
old
server,
it'll
fall
back
on
the
extensions
kind.
It'll
just
find
like
what
first
was
the
most
recent
of.
A
A
B
B
A
A
A
B
Okay
and
then
what
about
breaking
out
like
like
Federation
commands
are
now
built
into
cube
control
right
now,
I
think
you
know
we
need
to.
We
need
to
cut
the
cord
a
little
bit
on
Federation
to
with
that,
like
with
that
simple
I
think
if
we
sort
of
make
that
beat
the
Federation
folks
problem
to
do
their
own
extension,
yeah.
B
Know
and
we
were
doing
cube
admin
we
sort
of
did
it
as
a
separate
binary
with
the
idea
that,
like
okay
as
the
extension
stuff
comes
in,
maybe
now
we
have
to
control
admin,
but
that
would
be
an
extension.
And
so
we
try
to
keep
that
somewhat
clean
from
from
the
cube
control
stuff.
So,
okay,
really
we're
just
I.
B
And
as
we
actually
built
it
out
in
in
the
repo
we
tried
not
to
actually
we
try
and
have
a
clean
set
of
dependencies
to
the
rest
of
the
repo.
So
we
should
have
a
relatively
painless
extraction
for
cube,
Natalie's
you're,
going
to
have
to
deal
with
like
the
other
problems,
which
is
integration
and
testing
yeah.
B
A
B
Okay,
so
I
mean
what
do
you
think
so
I
mean
like?
Are
there
sort
of
smaller
milestones
that
we
can
actually
start
chipping
away
at
that?
We
can
sort
of
declare
victory
on
that
we
can
motivate
people
around
yeah.
A
B
B
That
des
would
be
more
like
traditional
package
management,
where
you
specify
the
versions
that
you
can
work
with.
You
do
simpler,
and
then
you
know
you
essentially,
you
know
have
a
way
that
you
sort
of
resolve
all
that
bring
it
together
pick
the
agreed-upon
version
that
works
for
everybody
and
then
lock
it
down.
Yeah.
A
A
B
One
of
the
things
and
and
I
don't
know
and
I
think
I'd
love
to
get
your
opinion
on
this,
because
I
think
it's
going
to
impact
this
group.
A
lot
is
like
there
is
like
all
the
package
utils
stuff
right,
yeah,
Gumble
time,
though,
like
we
have
this
goal,
where
we
want
every
piece
of
code
to
roll
up
to
a
sig
I
have
no
idea
what
sig
package
HL
ends
up
in
right.
B
A
B
Yeah
no
I
I,
don't
know
I
mean
just
think
about,
like
I,
think
we're
going
to
be
forcing
this
together
for
this
than
anybody
else's,
because
it's
the
sort
of
a
night
to
have
for
everybody
else.
It's
a
requirement
as
to
control,
gets
drawn
out
and
and
nobody's
in
a
hurry
to
actually
give
all
the
package
util
stuff.
A
home
I
suggested
that
we
form
an
architecture
sig
and
we
put
it
there
because
it
actually
is
kind
of
a
core
thing,
and
so,
like
you
know,
it's
part
of
sort
of
rolling
up
your
sleeves.
D
A
B
B
Kind
of
a
mess-
okay,
but
you
know
because
this
you
know
this
group
is
a
forcing
function.
If
we
have
strong
opinions
on
who
the
owner
should
be
like
we
can,
we
can
help
push
that
and
make
it
out.
Yeah.
B
All
right,
I,
don't
know:
okay,
let's
let's
table
that.
But
but
if
you
see,
if
you
see
obvious
candidates
or
EPC
discussions,
we
can
help
connect
the
dots
their
enjoyment
for
we're
here
for
drugger.
It.
A
Sounds
good
something
to
think
about
over
the
next
weeks
and
months,
great
okay,
we
have
two
proposals:
the
Alpha
command
and
the
unset
command.
We
try
and
the
reason
for
the
proposals
here
is
just
to
try
and
get
a
forcing
function
on
resolution.
For
some
of
these
things,
the
first
one
it
looks
like
a
PR
for
the
code,
so
I'm
not
sure
this
is
typically
I.
Think
these
are
for
a
proposal
on
whether
or
not
to
do
something
you're,
not
acceptance
of
code,
but
we
can.
A
C
B
I
mean
so
first
of
all,
I
think
you
know
my
gut
is
and
I
you
know
and
looking
at
it
from
the
outside.
G-Cloud
has
sort
of
an
alpha
group
and
a
beta
group,
for
you
know
Alpha
and
Beta
commands
there's.
You
know
it's
very
clear
that
you're
running
something
that's
not
going
to
be
long-term,
supported
and
seems
to
be
working
pretty
well.
We
have
an
alpha
group
in
cube
admin
and
we
do
it
as
a
way
to
just
sort
of
say,
hey
look.
B
C
A
C
B
Like
I
said,
I
think
it's
probably
just
borrowing
the
pattern
from
from
g-cloud
and
all
indications
that
I've
seen
is
that
that's
worked
out.
Okay,
yeah.
C
B
A
B
A
The
maybe
in
a
couple
weeks
or
the
next
meeting,
we
can
come
back
to
the
debug
one
Fabiana
DIGI+
proposals
here
or
who
added
them.
Yeah.
C
A
B
B
C
Exactly
I
think
that
that's
probably
the
only
other
option,
so
it's
it's
actually
not
a
matter
of
what's
the
name,
but
we
either
have
a
separate
command
like
I'm
set
or
do
something
under
set.
We
use
that
kind
of
pattern
in
some
other
places
like
set
school
year
or
that
should
remove
something
like
that.
I
personally,
don't
like
that
too
much.
But
so
I
would
personally
be
in
favor
of
a
separate
two
men,
so
probably
unset,
but.
A
C
Yeah
my
major
concern
with
you
know:
there's
such
clear
or
something
like
that
under
set
is
the
mutually
exclusive
problem
of
flags.
You
know
because
sometimes
set
has
a
lot
of
flags
and
they
really
don't
you
know
conform
with
clear,
so
clear
is
absolutely
exclusive.
With
most
of
the
other
flags
you
have
under
set.
So
that's
kind
of
automatically
for
me
makes
it
a
separate
command,
but
it's
not
a
it's,
not
a
huge
issue.
C
If
you
don't
have
too
many
flags
under
your
set
command,
but
you
know,
if
you
start
getting
a
lot,
it
gets
complicated
right,
but
anyway,
I
really
didn't
check
other
places
where
we
use
that
pattern.
If
we
are
broadly
using
those
kind
of
things
that
could
be
an
argument
in
favor
of
that's
as
clear
or
something
like
that,
I.
B
B
Program
results
in
essentially
a
full
package
deployment
of
all
the
objects
that
you
want
to
have
happen,
and
ideally
this
would
be
something
that
we
could
pass
in
pipe
to
apply
across
the
entire
set
of
objects,
one
of
the
problems
with
prune
and
why
everybody's
afraid
to
actually
enable
it
is
that
if
you
get
the
labels
mixed
up
wrong
with
the
set
of
objects,
you
end
up
the
leading
a
bunch
of
stuff.
Do
we
have
essentially
an
object
that
I
can
pipe
to
apply
prune?
B
B
What
I
guess
the
question
is
that
can
I
actually
generate
something
pipe
it
to
tube
control,
applied
prune
and
have
a
depose
all
the
objects
and
a
set
of
labels
that
we're
using
to
actually
prove
right,
because
right
now,
I
think
the
problem
is:
is
that
the
set
of
objects
that
you're
actually
creating
and
then
the
labels
that
you're
using
for
prune
those
things
come
from
two
different
places
right?
Some
of
them
come
from
the
file
system
maximum
come
in
like
if
we
actually
had
one
thing
that
we
could
pipe
all
of
a
sudden.
A
I
totally
agree
with
what
you're
saying
I
think
it's
super
dangerous
I
agree
with
why
it's
dangerous
one
I
hadn't
been
thinking
about
that
particular
use
case.
The
way
I've
been
thinking
about
it
is
another
thing
is
when
you
run
so
when
you
want
to
fly
on
a
directory.
Another
fear
I
have
is,
if
you
run
it
on
a
subdirectory.
Even
if
you
get
the
labels
right,
you
guys
won't
be
on
all
part
of
your
tree
and
you
just
blow
away
a
bunch
of
stuff.
Oh
yeah
right!
That's.
B
It's
a
pretty
sharp
knife,
so
I'm
thinking
can
we
come
up
and
I'm
looking
at
from
the
point
of
view
of
Kasaan
it
with
K
cyan't
we're
going
to
have
a
way
to
actually
bring
together
a
whole
looking
stuff
and
generate
it.
As
we
generate
the
thing
with
the
labels,
it's
going
to
be
a
lot
harder
to
screw
that
up,
yeah
yeah,
there's
any
prior
art
there
I
want
to
make
sure
I
look
at
it.
B
A
An
issue
for
this
all
throw
it
in
the
notes,
and
it
goes
at
it
from
the
angle
where
and
I
think
these
could
work
together,
like
I,
want
with
prune
a
like
a
structured
object
that
has
the
labels
yeah
and
I
was
thinking
that
that
object
would
live
at
the
root
of
the
director
you'd
apply,
and
so,
when
you,
when
apply
it'll
check
for
that
that
route
and
if
it
doesn't
exist
there,
it's
not
going
to
apply
because
then
it
solves
that
other
problem.
But
we
could
also
probably
accept
it
on
standard
in
yeah.
B
Well,
it's
like
right
now
in
standard
you
can
take
a
list
and
then
and
then
to
control,
knows
how
to
like
parse
that
up
and
create
across
the
entire
thing.
I
know
well,
okay!
So,
let's,
let's
talk
to
take
that
offline
I
just
wanted
to
just
want
to
do
a
quick
question
on
that
and
see
if
there
was
anything
that
I
hadn't
seen
you
know.
C
Right
thank
enjoy
them.
Thank
you
feel
yeah.
We
are
actually
out
of
time
already.
So
any
wrap-up
you
guys
when
I
do.
Otherwise
we
are
done
here-
we're
probably
going
to
have
another
CLI
meeting
next
week
because
we
can't
fold
last
week
so,
but
you
know
a
lot
of
stuff
to
discuss
all
next
week
is
game
for
for
meeting
so
na
wrap-ups.
I
probably
won't
be
here
next
week.