►
From YouTube: Kubernetes SIG CLI 20190814
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
Okay,
good
morning,
good
afternoon,
good
evening,
depending
on
where
you
are
it's
August
14
2019-
and
this
is
another
of
our
bi-weekly
six
year
live
meetings.
My
name
is
Matty
and
I'll,
be
your
host
today,
announcements
all
on
a
thing,
Phil
and
Shawn
for
filling
in
the
past
month,
when
I
was
gone,
they
did
an
awesome
job.
I
see
both
recordings
online
I
see
agenda,
updated,
I
haven't
got
through
all
of
that,
yet
it's
still
a
second
week
since
I
returned.
A
B
I,
don't
want
too
much
and
that's
Thursday.
The
27th
is
code
freeze,
so
we've
got
what
about
a
week
and
a
half
a
little
bit
close
to
two
weeks
until
code
freeze
and
then
just
one
other
quick
note
for
the
last
couple
meetings
we've
been
mentioning,
even
though
it's
the
long
way
away.
Coupe
con
in
San
Diego
we're
hoping
to
have
a
face-to-face
and
actually
put
some
faces
to
names
and
get
the
six
CLI
community
together
there,
and
that
is
November
18th
through
the
21st
in
San
Diego.
A
Which
also
reminds
me,
I,
remember
talking
with
Phil.
Sean
was
also
asking
me
about
the
same
topic:
how
the
majority
of
people
are
feeling
about
us,
because
we
are
as
all
the
six
we
can
apply
for
spots
in
the
agenda
for
holding
six
Eli
designated
meetings.
We
have
basically
two
options,
which
is
what
we
were
doing
up
till
now,
having
two
meetings
where
one
is
devoted
to
introductions
and
the
other
one
is
devoted
to
deep
dives
for
60
Li,
but
actually
Phil
and
I.
A
We
discussed
this
last
time
in
Copenhagen
and
we
would
like
to
give
it
a
try
to
a
different
approach
this
time.
So,
instead
of
having
two
smaller
meetings,
we
would
prefer
having
a
one
combined
longer
meeting,
which
will
allow
us
to
discuss
more
in
depth
about
the
directions,
maybe
holding
some
kind
of
a
panel
discussions
Q&A
with
people
attending
that
are
interested
in
contributing
or
the
changes
are
happening
in
in
in
60
Li.
So
I'm
more
asking
about
your
input.
What
you
think
about
the
the
proposed
approach.
A
A
Meet
after
a
different
but
meet
on
with
some
beverages
and
our
heads,
okay,
I
will
that
I,
like
I,
said
before
I'm
kind
of
new
for
I
was
gone
for
over
a
month,
so
I
don't
know.
What's
the
test
great
at
this
point
in
time,
and
what
is
the
test
coverage
so
I'll?
Let
Sean
give
an
update
or
he
will
know
best
who
was
the
person
looking
after
CI
in
the
past
of
the
weeks.
B
The
most
of
the
tests
are
green.
Some
of
the
gke
google
specific
tests
trying
to
bring
up
the
clusters
at
head
is
actually
having
problems
just
for
google
and
I'm
looking
into
that.
So
we
we
had
solved
one
specific
group,
one
specific
Google,
one
Google
specific
problem
and,
and
then
we
uncovered
another
so
so
I'll
take
that
on
as
an
action
item
to
to
get
that
back
to
green
okay.
Thanks
for
the
update.
A
A
quick
reminder:
we
have
a
remember
whether
that
was
that
it
is
we
have
not
updated
list
of.
If
you
want
to
sign
up
and
be
our
six
Eli
on
call
feel
free
to
add
your
name
for
the
next
period,
I'm
guessing
that
Sean
will
be
the
one
that
I'll
put
in
there
for
the
past
time
and
hopefully
for
the
next
week,
I'm
promised
I'm
gonna.
Take
that
one
over
after
our
next
call.
A
C
C
The
other
thing
that
the
conformance
working
group
has
been
trying
to
do
is
actually
collapse.
The
number
of
images
required
to
run
a
test
pass
at
one
point:
they
were
up
close
to
I
think
around
30
different
images
and
several
gigabytes
of
test
content,
and
a
lot
of
those
images
were
only
used
once
and
you
know.
C
Basically,
the
content
of
what's
running
in
that
container
doesn't
really
change
the
way
kubernetes
runs
and
so
they've
been
trying
to
narrow
that
down,
so
that
it
takes
less
download
time,
it's
more
achievable
for
somebody
to
be
able
to
run
this
offline
by
playing
to
their
own
registry.
So
that's
just
kind
of
a
bit
of
background
on
this.
A
I'm,
just
looking
into
the
code
to
answer
your
question,
but
from
a
quick
look
and
especially
that
it
lives
inside
of
keeps
ETL
cubes,
you
filled
that
go
all
the
signs
are
pointing
me
that
this
is
our
our.
As
in
six,
you
I
own
tests,
which
basically
means
we
are
free
to
change
it.
Okay,
that
means
basically
picking
something
other
than
that
is
honestly
I,
don't
see
any
objections.
A
A
C
I'm,
proposing
is
no
database
at
all
here,
which
has
there
are
other
test
cases
out
there,
that
test
things
like
services
and
load
balancers,
and
so,
if
the
goal
is
to
test
the
cube
control,
behavior
I'd
rather
try
to
narrow
this
down
in
focus
so
that
it
runs
faster
and
is
easier
to
maintain
rather
than
it
depending
on.
You
know:
five
different
deployments,
yeah.
D
B
D
C
D
Need
to
send
that
it's
sending
the
right
responses
to
the
API
server
and
get
e
in
reading
parsing.
Those
responses
appropriately
right,
yeah
and
it'd,
be
much
more
desirable
for
us
to
you
know,
have
a
test
that
could
run
in
minutes
rather
than
hours
as
part
of
nice
tweet,
or
something
like
that
and
could
just
never.
B
D
And
we
do
have
a
test
framework
for
that
now,
which
brings
up
e
VI
server
a
net
CD
in
but
doesn't
bring
up
the
whole
cluster
right.
We
can
just
poke
data
in
there
and
Shawn's
been
looking
at
that
in
as
part
of
the
work
to
move
coop
control
out
of
to
a
nice
cover
netis.
The
hope
is
that,
when
it's
in
its
own
repo,
we
will
like
the
majority
of
the
tests,
will
be
a
back
form.
Okay,.
C
And
I
think
that's
absolutely
the
right
thing
to
do
to
increase
coverage
of
kube
control
itself,
especially
as
you're
adding
new
features,
but
that
sticking
point
on
this
one
is
that
it's
currently
marked
conformance.
And
so,
if
the
goal
is
to
move
things
out
of
the
Cooper
neighs
repo,
that
means
that
they
probably
can't
be
conformance
tests.
And
so,
if
you're,
suggesting
that
we
rip
this
out
of
conformance,
that's
a
little
bit
different
thing
than
updating
it
in
place.
You
have
a
preference
on
which
direction
they're.
So
these
ooh.
C
D
Think
is
what
you
know.
Ma
che
was
saying
that
may
be.
The
right
thing
to
do
is
just
remove
conformance
for
it
or
if
we
can't
remove
conformance
for
it
from
it.
It
means
that
it's
probably
testing
something
else,
even
though
it
happens
to
be
in
the
C
apply
repo,
and
if
it
is,
then
we
should
probably
figure
out
what
it
is
and
maybe
move
the
test.
Someplace
else,
yeah.
C
And
so
so,
if
you
don't
think,
Cube
control
needs
to
be
in
conformance
one
direction
is
we
could
take
that
off
and
then,
if
there's
other
behaviors
that
are
outside
it
cube
control
that
want
to
be
tested,
then
I
would
probably
ask
someone
like
cig
apps
to
take
that
over
if
they're,
if
for
some
reason,
they
were
looking
at
that
as
a
signal
to
whether
or
not
deployments
succeed
and
do
the
right
thing
does
that
does
that
make
sense
exactly
yeah?
That
makes
perfect
sense.
D
A
A
That's
definitely
not
part
of
the
keep
CTL,
because
keep
see
data
created,
I'm,
pretty
sure
that
it
is
being
tested
thousand
times
across
the
entire
ete
and
should
definitely
not
be
part
of
conformance
I'm
more
than
happy
to
drop
it,
as
one
reasoning
is
if
we
can
find
a
cig,
apps
scenario
covering
the
same
thing
just
say
yeah.
This
is
this
is
a
duplicate,
the
the
the
other
test,
and
we
can
just
remove
it
as
such,
and
that
will
be
perfectly
fine
for
for
me
as
well.
D
A
A
A
A
A
C
Sounds
good,
so
what
I'll
do
is
I'll
get
a
PR
start
on
this
to
to
simply
remove
the
conformance
tag
for
this,
and
then
I'll
follow
up
with
sig
apps
to
see
if
they
want
to
to
make
sure
they're
comfortable
with
the
coverage
that
they
have
or
see
if
they
want
to
otherwise
adopt
this
test
case,
and
do
it
in
a
different
way.
I
mean.
A
I
would
check
whether
that
there
isn't
already
a
test
yeah
well
torture,
first,
which
is
like
identical
or
close
identical
to
that,
and
if
that's
the
case,
then
you
know,
instead
of
just
dropping
the
conformance
I
would
do
both
dropping
the
conformance
and
the
ties
itself
stating
clearly.
This
is
a
clear
duplicate
of
the
other
one
and
eventually
bumping
the
other
one
to
a
confirm
this
level.
C
A
B
Okay,
so
yeah
sorry,
just
one
quick
mention
we
once
we
moved
to
the
second
phase
of
our
new
coop
cuddle
repo.
You
know
the
first
phase,
having
moved
our
code
to
staging
we're,
going
to
be
looking
to
to
create
this
new
integration
test
framework.
Actually,
we
already
have
integration
test
framework
and
we're
going
to
need
help
actually
to
fill
out
the
test
cases
for
that
integration
test
framework.
It's
not
it's
not
on
the
horizon
within
the
next
week
or
two,
but
it
will
be
coming
up
for
anyone
who
wants
to
contribute.
A
E
So
in
the
past
few
weeks
we
discussed
that
we
want
to
move
the
code.
Although
the
binary
and
libraries
we
developed
in
C
our
experimental,
including
apply,
including
spatters,
including
prune
and
dynamic
commands
that
phil
has
developed
into
CRI
u2's.
So
I
have
moved
the
code
there,
but
now
we
need
to
give
a
good
name
for
this
binary.
So
this
binary
currently
includes
several
sub
comments
like
apply
prune
staggers,
and
we
have
I
discussed
with
char
and
feel
about
some
candidates.
Names
like
K
you'd
use,
k
zr.
E
A
If
so,
and
and
that
also
follows
what
what
you
just
said-
that
phil
suggested
is
having
wine
one
binary,
/
/
function,
/
functionality.
In
that
case
the
naming
scheme
would
be
cube,
CTL,
whatever
the
functionality
is
so
we
would
use
cube.
Ctl
applied,
maybe
not
necessarily
apply,
but
maybe
apply
to
or
apply
v2
or
something
like
that.
But
with
cube
CTL
as
a
prefix,
then
each
and
every
single
cube
serial
user
can
grab
those
experimental
commands
and
and
use
them
locally,
as
as
a
plug-in
and
and
test
and
report
back
to
us.
A
The
experimental
commands
even
more
but
I'm
I
think
we
would
have
to
figure
out
with
Kern,
Elias
and
Ahmed
how
we
could
publish
the
experimental
commands
as
part
of
the
crew
index
so
that
people
that
will
be
using
the
QCD
of
plug-ins
search
functionality
from
crew
will
will
be
able
to
find
the
experimental
commands
that
we
provide,
but
with
a
clear
information
that
this
is
an
experimental
experimental
command.
But
at
the
same
time
both
of
these
will
expose
the
kinetic
experimental
events
to
a
wider
audience.
That's
how
others
are
feeling
about
it.
I.
D
Think
some
of
the
commands
we
should
probably
partition
like
what
are?
We
really
are
experimental
versus
what
are
just
more
mature
versions.
Actually,
a
commands
already
have
like
the
the
intent
with
the
apply
command.
I
think
is
actually
to
be
a
more
focused
which
perversion
of
apply
where
it
has
like
declarative
deletion,
for
instance,
where
you
can
annotate
a
resource
to
be
deleted,
and
so
now
it
actually
has
a
much
more
mature
way
of
managing
deletion
of
resources
in
status.
Similarly,
as
I'm
the.
E
D
Connotations
of
experimental,
like
these
things,
are,
have
less
mileage
on
them
right
and
some
of
the
ideas
are
newer.
So
in
that
sense
we
do
want
the
feedback
and
they're
more
experimental
from
the
perspective
of
they're,
just
they've
been
useless
and
and
some
of
the
ideas
we
need
to
get
feedback
on,
but
they
they
aren't
experimental
in
the
sense
that
we
are
planning
on
abandoning
them
or
something
like
that
without
without
wrapping
them
into
the
maintenance,
then
the
diamond,
the
name
dynamic
plans
actually
are
relatively
experimental
in
that
we
don't
know
where
you
know.
D
B
D
D
Yeah
I,
don't
know
like,
and
do
we
want
like,
so
one
option
is
just
straight:
distribute
them
as
to
control,
plugins,
distribute
them
and
also
distribute
them,
not
as
food
control,
plugins
and
just
have
another
name
that
they
reside
under
that
way,
you
don't
need
to
run
run
them
through,
like
a
symbolic
link.
Thing
doesn't
work
well
if
we
actually
just
want
to
set
it
commands
directly.
So.
E
A
A
A
Combination
of
all
the
toys
that
is
possible
will
be
only
slight
of
a
burden
eventually,
eventually,
we
can
consider
having
an
image
with
all
or
n
couple
of
variations
that
will
not
include
the
experimental
commands,
so
one
will
be
with
everything
one
will
be
with
the
stable
ones
and
a
separate
with,
but
maybe
not
necessarily
separate
for
experimental,
but
have
different
levels
of
trust
against
users.
So
if
I
don't
trust,
my
users
I
only
allow
them
to
run
stable
if
I
trust
my
ears
or
they
can
use
both
stable
and
experimental.
A
One
does
not
oppose
the
other
one.
The
publishing
asked
using
a
skip,
CTO
plugins
will
definitely
provide
us
a
little
bit
more
visibility.
If
we
have
a
lot
of
crew
users
now
and
I
quickly
went
through
crew
index,
there
are
options
to
say
that
this
is
an
alpha
feature,
alpha
functionality
within
a
version
field,
and
that
basically
means
yeah.
We
will
say
it's
an
alpha
command
use
it
at
your
own
risk.
That's
it
and
and
stable
will
be
a
properly
version
that
that's
such.
F
A
A
B
A
D
A
Right
so
I
remember
the
APR,
where
someone
was
changing.
A
How
we
invoke
the
plugins,
with
my
review
on
that
PR
was
that
I
don't
want
I
want
to
maintain
the
functionality
of
replacing
a
process,
at
least
in
Linux
or
UNIX
systems.
Yes,
and
only
do
that,
the
the
alternate
approach
on
Windows,
unfortunately,
I
have
not
returned
to
the
PR
that
was
doing
this
I'll
try
to
find
it
up
and
see
whether
it
was
updated
or
sync
with
the
order,
if
not
how
we
submit
the
PR,
with
with
the
alternative
method
for
running
those
on
Windows
only
because
it
is
definitely
worth
fixing.
D
Yes,
I
agree:
if
we
do
this
as
a
plug-in
exercise,
I
kind
of
think
it
would
be
better
to
do
it.
As
a
single
point
like
to
solve
somewhat
like,
we
could
do
all
the
symbolic
linking
stuff
which
I
but
then
there's
still
a
namespace
collision
for
like
our
PI
commands
and
there's
one
we
rename
everything
to
but
like
if
it
was
just
under
a
new
sub
command,
which
was
itself
a
plug-in.
It
solves
a
lot
of
problems.
I
mean
we.
A
Could
do
both
the
reason
that
would
go
with
single
plugins
is
that
it
allows
users
to
be
more
picky
about
what
I
want
the
users,
none
necessarily
might
be
interested
in
everything,
yeah
and
I
think
we
can
build
yeah
in
case
of
the
image.
It
is
pretty
obvious.
Yes,
you
want
to
provide
all
of
them
at
once,
because,
like
I
said
before,
we
don't
want
to
multiply
our
options
here,
but.
D
The
default
case
or
the
case
where
you
don't
actually
want
koop
control
right
like
I,
can
imagine
I
just
want
this
one
command,
I,
don't
want
to
control
I,
don't
want
any
other.
Exactly
no
building
I
think
the
plan
is
to
build
every
single
and
as
its
own
binary,
that
command
will
probably
be
the
full
name,
so
it'd
be
like
still
be
coop
control,
prefix.
Whatever
the
namespace
is
command,
then
they
can
just
download
that
command.
C
E
A
F
C
F
B
F
A
F
D
D
The
original
purpose
of
this
repo
was
to
have
individual
binaries
for
libraries
published
in
CLI
runtime.
So
that's
consistent.
We
still
get
back,
give
individual
binaries.
Then
we
distribute
this
as
a
crew
plug-in
for
coop
control.
We
also
distribute
it
as
its
own
binary.
If
you
just
want
one,
if
you
just
effectively
want
apply
in
status.
E
D
D
Maybe
that's
it
right
and
you
can
actually
do
that
we
can
ship
the
same
binary.
Actually,
we
can
just
have
a
build
time
like
which
enables
that
class
as
well
the
same
way.
We
do
it
same
way.
We
inject
the
version
info
into
the
binary
when
we
build
it.
Okay,
selectively,
it's
like
feature
gating
right,
like
beat
you're
getting
miners
yeah.
A
D
Cool
all
right
so
we'll
just
follow
up
with
renaming
yeah,
okay,
perfect.
A
E
Can't
do
a
quick
stand-up
for
customized,
so
we
released
three
dot
1.0
several
weeks
ago
for
customized.
It
includes
the
extended
pad
feature
with
one
hand:
you
can
modify
multiple
objects
so
that
one
was
included
and
it
received
very
good
feedback
after
people
start
to
use
it
and
for
in
the
past
two
weeks
we
don't
have
much
updates
for
customized,
but
in
a
coming
release,
probably
this
month
the
inland
patch
will
be
included
so
I
have
implement
that.
But
it's
now
it's
only
our
hats,
but
not
released.
Yet.
C
A
Kuya
is
is
saying
that
they
are
improving
stability
and
test
coverage
and
reach
500
starts
definitely
Congrats.
They
congrats
on
on
that
milestone.
Okay,
I
think
that's
pretty
much
all
for
for
today.
Thank
you
very
much
for
your
participation
and
today
discussions
and
see
you
in
two
weeks.
Thank
you.
Bye,
yeah,.