►
From YouTube: Kubernetes SIG CLI 20190213
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
today,
is
February
13th,
and
this
is
our
another
sexy
like
bi-weekly
meeting.
Let
me
quickly
share
my
screen
and
hope
that
yeah,
that
the
binary
or
okay,
you
should
see
the
the
agenda.
Hopefully
so
I
don't
have
any
particular
announcements.
C
B
A
B
We
immediately
had
one
issue
with
the
tape
scream,
which
was
on
the
failing
test
from
AWS
I
later
figured
out
that
it
was
a
funding
issue
and
someone
sppf,
someone
who
does
handle
had
created
and
someone
created
a
an
issue
for
that
and
I,
don't
know
when
it
would
get
fixed
but
I'm
still
watching,
and
that
particular
test
is
still
failing
to
date.
A
Perfect
thanks
very
much
the
next
one
call
will
be
shown
and
quick
reminder
if
you're
willing
to
sign
up
for
the
uncle
test
on
control.
There
is
a
sexy
like
testing
playbook
document
we
have
spots,
starting
from
from
the
next
find
slot
over
through
March
April
and
further
feel
free
to
add
yourself
there.
If
you
have
any
questions
with
regards,
what
are
the
duties,
how
it
works?
A
D
D
D
D
So
so
the
final
state
on
the
cluster
is
we
have
objective,
foo
and
upgrade
paths
which
which
can
match
the
directory
files,
all
the
configuration
in
the
directories-
and
this
is
the
motivation-
and
this
is
the
workflow
we
want
to
support
by
cube,
cut
or
applied
prune.
And
now,
let's
see
what
is
the
current
state,
and
this
is
the
Alpha
prune
it
is.
D
It
supports
a
L
which
is
a
label
selector,
and
when
you
give
that
label
selector,
it
will
curry
all
the
objects
cross
our
resources
and
namespaces
to
find
the
matching
ones
and
amount
of
matching
ones.
It
will
select
the
objects
that
has
the
last
apply
annotations
so
that
those
objects
are
created
by
previous
apply
and
then
for
any
objects.
That
is
with
guards.
Let's
apply
configuration
but
not
shown
up
in
the
new
apply
set.
D
Those
objects
will
be
deleted,
so
the
limit
mainly
from
two
parts
one,
is
that
it's
not
efficient,
since
it
has
many
number
of
queries
and
second
one
is
pretty
risky,
so
users
have
to
be
really
careful
when
they
use
this
feature
to
make
sure
that
they
don't
delete
and
objects
and
intentionally,
let's
say
if
they
have
a
typo
in
this
command
line
or
this
label
is
this
label
has
some
overlaps
with
some
other
directories.
The
objects
on
the
cluster
might
be
deleted
without
user
attention.
So
we
don't
want
that
that
kind
of
disaster
happen.
D
So
this
is
our
new
approach.
It
is
based
on
concept
called
apply,
config
object,
so
here
is
an
example
of
such
object.
It
has
a
label
which
is
cube,
Cottle,
kubernetes,
dot,
io
/
apply
config,
oops,
it's
true
and
based
on
this
label.
We
say
this
object
is
a
configure
ply
object
and
it
also
has
the
information
of
a
list
of
objects
that
will
be
applied
to
the
cluster.
So
here
we
call
that
list
of
objects
is
saved
in
the
last
apply
objects
annotation.
D
Here,
for
example,
we
have
three
objects
which
are
to
be
applied
to
a
cluster,
and
something
we
need
to
note
is
that
this
apply.
Config
object
is
not
generated
by
cube
title
itself.
It
should
be
generated
by
other
tools
such
as
customize,
our
helm
or
even
add
other
tools
on
user
can
created
this
object
manually.
D
They
just
need
ask
this
label
and
add
this
annotation
correctly
to
match
whatever
resource
is
present
in
files
or
directories.
Ok,
so
here,
let's
see
with
space,
apply
config
object,
the
new
workflow.
So,
let's
say
in
a
directory:
I
have
a
apply.
Config
object,
a
healthful
bar
the
first
time.
I
run
apply
for
this
whole
directory.
So
cute
cattle
will
see,
there
is
a
conflict.
Apply
config
object
by
by
this
label
apply
a
conflict
Troop
at
this
level
and
then,
since
the
clusters
is
now
is
intense.
D
The
first
time
I
was
trying
base
directory,
so
it
couldn't
find
an
ultra
configure
in
a
cluster,
so
it
will
cube.
Cattle
will
go
ahead
and
create
that
object
as
well
as
full
bar
in
a
cluster.
And
now
let's
say
we,
the
director
is
modified.
We
remove
the
bar
and
we
add
bath.
At
the
same
time,
the
applied
config
object
and
need
to
be
updated.
So,
as
you
can
see
as
we
now,
we
have
full
bar
in
the
annotation.
Sorry,
we
have
footbaths
in
the
annotation
instead
of
full
bar.
D
So
this
time
when
queue
ketosis
the
the
directory,
it
will
first
effect
the
apply.
Config
object
from
the
cluster,
which
is
the
yellow
one.
If
you
look
at
that's
the
life
object
for
apply
config,
so
there
the
last
objects
has
the
least
at
fubar,
but
in
the
new
of
psychic
object
we
have
the
list
of
objects
as
fruit
bats,
so
those
two
sets
are
different
and
cue.
Cattle
will
compute.
D
The
difference
of
those
two
sets
and
create
the
new
creates
the
new
resource
which
is
passed
and
deletes
the
the
old
resources
which
is
our
okay
and
then
updates
the
cluster
okay.
So
this
is
the
the
state
of
the
classroom
we
have,
who
may
be
updated,
but
bar
is
deleted
and
bass
is
created
and
apply.
Config
object
only
in
the
cluster
is
also
updated,
and
now
this
live
object
has
the
same
content
as
the
apply
config
object
from
the
directory.
E
D
C
E
D
D
D
First
I
use
data
to
save
the
list
of
resources,
and
then
we
realized
there
might
be
some
other
cases
we
want
to
support.
For
example,
people
may
want
to
use
other
CRD
kind
of
resource
as
the
apply
config
object
and
if
we
allowed
to
add
annotation
as
the
annotation,
we
can
add
the
list
of
resources,
then
other
types
will
be
also
supported,
so
this
apply.
D
Config
object
is
not
limited
to
only
config
map
kind,
so
other
kind
of
objects
can
also
be
recognized
as
apply
config
object
as
long
as
they
have
this
label
and
the
list
of
resources
could
be
found
from
the
annotation.
If
we
save
that
list
of
resources
and
data
data
field
is
not
available
in
many
other
type
of
resources.
D
G
E
But
if
it
works
with
something
like
the
application
CRD,
then
that
is
a
TRD
and
controller,
and
a
lot
of
tools
can
then
go
use.
That
and
I
think
that
that
kind
of
cross
compatibility
is
important.
I
mean
pushing
things
like
this
server-side
I
also
think
add
to
it
that
this
kind
of
gets
into
a
little
bit
of
dependency
management
and
resolution
and
I
know
when
we
discussed
application
in
the
original
cap
for
it
before
we
were
doing
caps
for
everything-
and
this
was
talked
about.
E
E
D
I
think
the
dependency
management
is
a
different
set
of
problem
so
since
that
problem
could
be
handled
by
other
tools
and
Keuka,
dough
is
kind
of
not
on
that
kind
of
problem,
and
so
what
we
want
to
provide
by
this
new
Pro
approach
is
to
given
the
ability
for
other
tools
to
use
this
prong,
so
if
they
so
regardless
of
how
they
manage
the
dependency
of
how
they
generate
their
resources.
So
as
long
as
they
have
the
right,
if,
as
long
as
they
have
this
a
psychotic
object,
kill
Ketil
can
handle
the
pruning
action
correctly.
E
That
assumes
they
use
coop
cuddle
and
more
and
more
people
want
to
be
able
to
use
other
tools,
which
means
it
doesn't
work
in
or
operably.
It
always
assumes
coop
cuddle
is
here.
So
if
somebody
is
rotating
a
tool
that
uses
one
of
the
kubernetes
clients
like
the
JavaScript
client,
they
would
not
be
able
to
work
with
us
because
they
go
through
cuddle.
Yeah.
G
They
can't
work
that
now
the
the
the
prune
is
in
it's
currently
in
in
Koo
cuddle,
now,
just
just
to
get
back
to
a
little
bit
of
the
motivation
for
the
prune.
So
so
NEC
ICD
system
that
is
going
to
continually
apply
a
set
of
objects
is
gonna
need
a
prune
which
was
kind
of
the
motivation
and
why
it
was
put
into
koo
cuddle.
G
You
know
over
a
year
ago
the
current
as
Jinping
mentioned,
the
current
implementation
doesn't
work
so
I,
don't
think
the
question
is
whether
or
not
we're
going
to
put
prune
in,
because
I
think
that
you
know
that
it's
already
in
as
elf
alpha.
The
question
is
how
we're
going
to
approach
it
and
the
current.
As
we
mentioned,
the
current
implementation
doesn't
work.
G
C
There's
a
couple
like
ways
of
talking
about
this
like
one
is
applied,
needs
to
be
able
to
delete
objects.
I,
don't
think,
there's
any
debate
about
that.
It
can
create
update
objects,
but
it
can't
delete
them
seems
like
odd
place
to
be
in,
and
we
know
that
it.
There
was
the
intention
there
because
we
did
implement
it.
C
We
just
implemented
it
in
a
way
that
we
knew
was
bad,
but
we
had
to
do
it
as
an
expediency
for
plug-in
or
add-ons,
so
actually,
criminis
itself
uses
apply
to
manage
add-ons
and
uses
apply
prune
to
delete
them
just
in
a
way.
That's
not
so
great,
so
I
agree
with
this
silver
side
pieces
I
think
we
have
to
like.
Unfortunately,
the
kubernetes
api
doesn't
provide
like
many
there's
a
provide
a
great
way
of
doing
cross
resource
api's,
there's
not
a
lot
of
them.
C
E
Go
ahead,
I
agree
with
that:
Phil
I,
guess
what
I'm
thinking
of
is.
Is
there
an
approach
that
can
be
done
that
allows
for
tool,
interoperability
right,
so
helm
doesn't
use
coop
control,
and
so,
if
it
goes
to
do
stuff
and
it
does
pruning
and
now
it
starts
not
interoperating
with
this,
it's
an
interoperability
problem
and
so
I'm
just
approaching
my
thoughts
on
this.
In
that
way,
how
do
we
make
this?
Is
there
another
way
to
do
it?
That
is
better
for
interoperability.
C
So
there's
a
couple
ways
we
could
make
this
interoperable,
like
the
logic,
isn't
radical,
we're
talking
about
doing
a
set
difference
and
deleting
everything
from
one
set
right
read
it
read
a
config
map
given
two
config
maps
subtract
one
from
the
other
then
delete
those
objects
and
then
write
the
config
map
back
so
like
one
level
of
interoperability
could
be.
Here
is
the
approach
we
are
going
to
use
based
on
config
maps,
it's
kind
of
a
standard
approach.
The
code
is
pretty
small
anyone
and
then
we
could,
like
you
develop
per
language.
E
C
E
A
H
A
C
C
E
A
A
D
C
C
H
C
A
Would
like
generally
starting
this
discussion
as
improvement,
because
we
know
the
cookies
mechanism
is-
is
completely
broken,
starting
with
this
as
an
alpha,
getting
feedback
from
circa
architecture
and,
most
importantly,
take
API
machinery
and
then,
if
they
don't
object
or
they
like
the
idea
or
will
be
supportive
for
the
idea,
can
actually
you
know
and
migrated
to
pay
it
off
within
115,
because
if
I'm,
right
correctly,
that
was
the
targeted
version.
So
that
sounds
good
to
me.
Yeah.
E
I'd
also
like
to
run
it
through,
say,
gaps
will
actually
bring
it
in
and
discuss
it
in
a
meeting
in
there,
because
we
deal
with
all
of
the
workloads
and
the
applications
in
there
and
those
changes.
This
has
practical
implications
there
so
I'm
happy
to
take
it
over
there
to
see
whether
we
can
come
up
with
any
ideas
in.
C
C
E
C
I
mean
it
would
be
great,
like
I,
think
helm
is
written
and
go
I
think
there's
a
lot
of
tools
that
are
written
and
go
like
like
if
we
could
produces,
as
a
library,
at
least
to
the
tools
that
are
written
and
go
and
say,
look
it's
a
best.
Effort
attempt
to
at
providing
this
functionality
is
sort
of
compromised
might
be
a
good
way
to
go
and
we
might
be
able
to
like
I
know
how
implemented
boom.
C
C
D
A
F
Yeah
I
don't
have
any
slides
or
anything
just
the
general
idea
is
that
customized
used
to
have
the
ability
to
execute
commands
in
the
secret
generator,
and
you
could
like,
let's
say,
you're
using
stops
to
encrypt
some
files.
It
could
you
know
it
could
execute,
stops
and
decrypt
those
files
for
you
and
it
was
taken
out
because
of
basically
someone
someone
could
ship
a
customized,
remote
customized
set
up.
That
will
execute
some.
You
know
something
on
your
computer
and
you.
F
Sort
of
malicious
command,
so
the
idea
over
here
is
to
to
use,
go
plugins
to
supply
like
the
functionality
to
you
know,
talk
into
shops
or
you
know
any
other
secret
management
tool
and
it
would
kind
of
enforce
an
API
alright,
where
you
give
it
an
array
of
arguments,
and
it
gives
you
back
an
array
of
key
value
pairs
and
then
also
it
wouldn't.
You
wouldn't
build
a
ship,
a
plug-in
as
part
of
your
customization
files.
E
Yeah
I,
think
and
I'll
reiterate:
I
have
two
concerns
about
it.
The
one
I
did
earlier
but
I'll
put
on
again
on
the
record
here
is
the
window
support
the
plugins.
They
came
out
and
go
1.8
and
they
didn't
then,
and
they
still
don't
have
window
support
and
I
know.
While
we
don't
do
a
great
job
of
testing
with
Windows
today
we
do
have
a
fair
amount
of
people
who
do
use
it
with
windows
and
they're.
Probably
just
this
big
silent
group
that
stuff
works
well
enough.
E
My
other
concern
with
it
and
I
don't
know
that
this
has
been
fixed,
but
with
plug-in
development.
When
you
create
plugins
everybody
who
does
it
has
to
have
the
same
path
from
the
root
of
the
file
system
for
the
plug-in
to
work
so
I
like
I'm
on
a
Mac
and
if
I
have
slash
user,
slash,
my
user
name
slash,
go
slash
the
path
to
it
that
doesn't
work
because
slash
users
slash
my
user
name
is
unique
to
my
system,
and
that
was
one
of
the
other
problems
with
it.
E
So
if
I
pass
it
off
to
Phil
and
he's
on
a
Mac,
slash
user,
slash
he's
using
it's
a
different
path,
and
so
there
is
an
interoperability
and
development
problem
with
that
as
well.
That
is
it's
a
pain
point
when
you're
building
plugins
and
so
those
two
things
make
me
question
using
the
built
in
plug-in
support.
So.
F
E
But
but
people
tend
to
share
this
stuff
on
github
and
co-develop
it
right
and
yeah.
That
model
is
then
broken
because
you
have
to
have
the
same
from
root,
and
so
those
people
who
want
to
develop
and
open
source
their
plugins
likes
ops,
I,
wouldn't
expect
everybody
who
uses
ops
to
develop
a
plug-in
would
probably
expect
one
sauce
plug-in
heck.
E
E
I
everybody
who
did
it
would
have
to
do
all
their
development
and
docker,
then,
and
you'd
have
to
go
set
up
that
build
tool
chain.
Although
this
raises
the
barrier
to
entry,
which
I
think
is
one
of
the
reasons
plugins
have
not
really
taken
off
that
plugins
package,
all
right,
I'll
see
you
too
Phil
I
see
you
have
your
hand
up.
Yeah.
C
C
A
A
A
So
I
didn't
know:
I
I
definitely
think
that
the
go
plugins
is
something
it
will
be
solved.
Sooner
than
later.
That's
your
house
yeah
how
broad
the
go
developers
scope
is
on
Windows
and
how
fast
we
can
get
to
the
point
where
go
modules
will
be
supportable,
I
I,
just
I,
don't
want
to
stop
us
from
doing
stuff
the
right
way,
because
I
I
I'm
pretty
sure
this
is
the
right
way
to
do
stuff.
That's
how
I've
been
using
similar
I've
been
solving
similar
problems
in
the
past
just
by
using
a
library.
A
E
Yeah
so
the
go
plugins,
they
haven't
moved
a
lot
since
they
came
out
in
1.8,
and
so
that's
one
of
the
reasons
that
I
don't
know
that
I
would
expect
them
to
to
move
unless
somebody
else
has
some
other
information
on.
What's
going
on
over
there
I
just
haven't
seen
much
movement
on
it
and
a
bunch
of
us
had
talked
about.
How
do
we
do
this
and
solve
this
problem
before
we
looked
into
this
and,
quite
frankly,
it
doesn't
look
like
it's
moved
since
then
somebody
says
plug
that's
go
modules.
E
Go
modules
is
different
than
plugins,
so
folks
are
talking
about
go
modules,
that's
basically
dependency
management,
it's
a
replacement
for
Ford
app
or
go
to
apps
or
glide,
or
any
of
those
things.
It
doesn't
solve
your
plug-in
problem
of
having
a
separate
binary,
released.
It's
tangental
to
that
good.
G
D
E
Coop
control
plugins
because,
yes,
using
any
executable
in
the
path
I
think
it's
a
fine
method,
in
fact,
so
what
helm
did
because
I
think
what
coop
control
does
is
a
little
different?
What
helm
actually
has
inside
is
it
has
a
couple
of
hook
points
where
it
says
we
have
this
event,
we'll
look
at
the
metadata
from
that
event,
and
from
that
we'll
actually
know
which
external
binary
to
call
you
can
intercept
the
stuff
along
the
way,
and
so
you
could
write
those
in
any
language
and
one
of
them
for
Hellman's
downloaders.
E
E
H
Thanks
so
there's
a
lot
more
stuff
to
talk
about
a
meeting.
This
is
just
a
professional
cap
and
it's
meant
to
unblock
people
who
were
actually
using
the
exact
feature
that,
as
it
existed
in
customized
before
so
I
think
it'd
be
pretty
non-controversial
to
put
in
a
you
know
to
let
this
thing
go
to
implementable,
because
you
know
it
could
use
customized
plug
it
could
use
cou,
cuddle,
plugins
and
all
it
does.
The
scope
of
this
is
extremely
constrained.
H
Concerns
you're,
not
the
security
concern
and
the
other,
and
the
original
approach
was
the
customization
file
could
actually
have
a
command
in
it,
and
that
would
run
so.
The
command
was
textually
in
the
customization
file
that
you
could
download
from
the
internet.
In
this
case,
what
the
customization
file
would
specify
is
a
local
executable
like
a
local
plugin,
so
the
risk
would
be.
H
The
attack
would
be
is
that
you
would
provide
someone
else
with
a
customization
file
that
they'd
be
running
downloading
and
running
remotely,
and
then
the
the
the
victim
would
have
to
have
a
you
know
the
right
plug-in
installed
and
the
worst
thing
that
could
happen
is
that
would
provide
a
key
value
pair.
So
it's
not
running
arbitrary
commands.
It's
well.
The
person
who
installed
the
plugin
it's
up
to
them
not
to
install
some
evil
plug-in
I
mean
there's
only
so
much.
H
You
can
do
you
always
if
you
want
to
provide
somebody
with
the
ability
to
generate
a
secret.
Somehow
some
code
has
to
execute
and
the
person
who
is
doing
this
has
to
make
a
choice
as
to
what
code
they
install
this.
This
there's
no
more
security
risk
adhere
than
there
already
is
in
cucuta
plugins,
as
they
currently
stand.
Yeah.
E
I
think
just
to
quickly
summarize
what
I
heard
you
say
was
that
in
the
old
model,
the
customized
customized,
but
actually,
if
it
saw
the
file,
it
would
go
out
fetch
something
and
install
it,
and
so
somebody
may
not
realize
that
the
external
tool
is
doing
this
in
the
new
model.
Somebody
who
runs
this
has
to
opt
in
to
go,
get
a
plug-in
that
supports
it
and
the
metadata
comes.
H
E
H
H
E
A
You
spend
some
time
and
try
to
figure
out
what
is
the
current
state
of
go
plugins
on
Windows?
What
is
basically
the
go
plugin
state
eventually,
if
there
are
some
plans
for
for
future,
when
we
could
expect
some
some
solution
for
the
windows
problem,
also
that
you
at
least
have
a
better
picture
and
put
that
information
in
the
cap
towards
the
the
place
where
you
said
that
it
is,
it
is
currently
not
supported
on
Windows,
so
that
we
at
least
provide
information
what
it
is
or
how
far
it
is
from
being
supportable.
A
C
F
Not
broken
because
I
could
lock
into
a
previous
version,
but
I
can't
I
can't
upgrade
so
any
of
the
new
bug
fixes
or
you
know,
new
features,
I
can't
I
can't
use
them
right
now.
I
think
that's
like
the
case
for
anyone
who
is
using
commands,
which
was
like
a
big
like
call.
You
know
using
customized
in
the
first
place.
This.
C
F
Okay,
that's
super
helpful
I
think
like
helm,
has
something
like
that,
like
there's,
like
a
hook
where
you
could
decrypt
all
your
files
and
then
run
apply
and
then
run
another
tool
to
re-encrypt
everything
but
like
during
that
time,
it's
decrypted
on
disk
and
if
you're
running
it
on
like
a
CI
or
I,
don't
know
yeah.
That
makes
sense.
F
Okay,
that
doesn't
really
bother
me
that
much
I
just
want
to
I
want
to
just
be
able
to
use
customize
build
and
have
everything
like
declared
in
the
file
and
not
have
to
I.
Don't
wanna
have
to
like
do
something
else
to
make
customized
work
right.
You
know
you
don't
want
to
go
back
to
writing
a
script
yeah.
C
A
A
That's
that's
also
possible.
Yeah
I
was
actually
chatting
with
King
Frank
earlier
about
the
cutting
the
functionality
of
the
customize
that
we
have
in
the
COA
runtime
just
to
the
core
functionality,
and
in
that
case,
customize
customize
itself
would
be
just
having
more
functionality
than
the
one
from
keeps
you
do.
Custom
entry
all
right,
awesome,
sorry,.
A
But
you
don't
need
to
do.
Mine.
I
was
actually
going
to
ask
Ahmed
about
the
plugin
manager,
because
I'm
pretty
sure
that
this
is
a
bigger
topic.
Are
you,
okay
with
us
pushing
that
to
the
next
meeting
in
two
weeks?
Yeah
I
think
it's
farm
cold.
That
will
be
that
be
awesome
and
I'm.
Pretty
sure
that
Phil
topic
can
be
also
moved
to
the
next
meeting.
A
I
He's
so
great
piyo
all
mottos
here
so
at
this
point
the
agenda
and
he's
the
information
about
the
request
and
the
Buck
O'neil
data
about
two
month
ago,
I
created
an
issue
for
keeps
TL
dump
command
and
I
also
submitted
requests
for
it,
which
is
solving
that
I
tried
to
get
some
attention
on
slack
channel,
but
I
wasn't
so
lucky,
and
maybe
it
doesn't
a
neutral.
If
you
exactly
the
keeps
TL
cluster
down
comment
with
an
Isis
flag,
then
only
the
list
of
the
nodes
are
done.
I
All
the
information
like
services
posts
are
not
available
and
it's
quite
simple
fix
and
it
was
introduced
by
mistake
during
the
one
of
them
factory
new
face.
So
I
also
added
that
proper
test
coverage
for
it
and
thanks
to
that,
we
never
regress
and
I
think
it's.
The
review
shouldn't
take
too
much
time,
but
I
don't
know
what
to
do
to
push
it
forward.
A.
A
A
A
Okay,
so
for
myself,
I'm
working
on
fixing
keep
CTL
explained,
that
is
with
regards
to
the
open,
API
VR
for
Sierra.
These
currently
open
cubes
ETL
explained
if
you
target
a
CR
D
from
the
PR
specifically
because
so
far
we
are
not
publishing
open
API
for
security.
So
if
you
do
use
CRT
and
the
explain
keeps
it
yellow
displaying
panics
I'm
trying
to
figure
out
what
is
actually
not
working
correctly,
there
I
think
I'm
I
have
a
working
solution
that
PR
it
will
be
actually
replaced
by
a
different
one,
which
will
have
a
proper
fix.