►
From YouTube: Kubernetes SIG CLI 20210224
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
evening,
good
afternoon,
depending
on
where
you
are
today
is
february
24th-
and
this
is
another
of
our
six
july
bi-weekly
meetings.
My
name
is
mate
and
I'll
be
your
host.
Today,
let
me
quickly
share
my
screen,
because
I
just
figured
out
that
I
haven't
done
it
yet
and
let
me
quickly
share
the
agenda
with
the
rest
of
the
group
and
the
chat,
and
since
I
was
talking
with
you
all,
I
didn't
open
all
the
windows.
A
Okay,
so
the
agenda
is
in
the
chat,
a
quick
announcement.
I
have
only
one
there's
a
code
freeze
in
less
than
two
weeks
on
march
9th.
A
So
if
you
have
anything
planned
for
121
make
sure
that
is
being
merged
by
that
date.
It
is
basically
less
than
two
weeks
away
from
from
today
sean
phil
eddie.
Do
we
have
any
additional
announcements.
B
Yeah
I
just
dropped
one
in
from
slack.
Zander
is
saying
that
there
is
a
future
blog
freeze.
So
if
we
want
to
write
a
blog
post
for
any
of
the
features
we
have,
the
due
date
is
march.
1St.
A
Yeah,
so
if
you
have
any
big
features
that
you're
working
on
for
121
make
sure
to
reach
out
to
xander
or
or
any
of
our
of
us
that
I
just
called
out
okay,
let's
see,
do
we
have
any
new
people
on
the
call,
I
think
we
have
feel
free
to
you
know,
introduce
yourself
tell
us
who
you
are
and
what
you're
expecting
to
to
get
from
six
eli.
C
Hey
yeah,
I'm
new
here
my
name
is
josh
curse.
I
just
I
sent
a
pull
request
for
a
feature
for
for
a
coupe
cuddle
top
and
and
it
was
asked
that
that
we
wanted
to
do
a
demo,
so
I
thought
you
know.
I
thought
I
didn't
even
know
that
you
guys
did
this,
so
I
thought
it
was
pretty
cool
and
you
know
here
I
am
to
to
do
it.
D
Okay,
I'm
not
new,
but
I've.
No,
I
don't
think
I've
ever
introduced
myself.
I
just
sort
of
show
up
and
I'm
really
quiet,
but
I'm
sally
and
I
used
to
be
on
mache's
team,
but
I've
recently
switched
teams,
but
I'm
still
very
much
interested
in
sig
cli.
So
I'm
here.
A
Do
we
have
anyone
else,
I'm
scrolling
through
the
participants
list,
and
I
think
that
I
more
or
less
am
familiar
with
the
rest
of
the
folks,
yeah,
okay
and
I'm
not
hearing
anyone
else
screaming
out
loud.
I
guess
that
everyone
that
wanted
to
speak
up
did
so.
A
C
C
Okay,
so
so
pretty
much,
I
I
do
a
lot
of
operations.
Work
kubernetes
is
the
central
to
that.
Oh
there's,
a
problem
that
I
I
commonly
face
where
we'll
have
nodes
that
that
are
out
of
resources-
and
you
run,
you
run
this
command
top
nodes
to
get
the
nodes
that
are
having
problems.
C
C
So
it's,
but
it's
actually
really
difficult
to
take
this
and
easily
just
say
which
pods
are
running
on
those
hosts
that
are
taking
up
all
the
resources,
and
so
there's
some
there's
some
things
you
can
do
to
to
get
it
to
work,
and
I
mean
it's
it's
possible,
but
I
just
wanted
to
make
it
really
easy
to
be
able
to
do
to
be
able
to
do
that.
C
And
so
that's
what
the
pull
requests
that
I
referenced
in
the
document
is
about
and
the
the
new
feature
is
this
node
name
flag
that
you
that
you
would
add
to
top
pods
and
it
puts
in
the
node
name
in
the
column
node
for
you,
and
that
gives
you
a
lot
of
possibilities
to
be
able
to
grep
for
a
node
that
you
know
has
problems
you
and
then
you
can
do
all
the
other
things
on
the
command
line
that
that
you
can
do
like
sort
and
whatever
on
on
cpu
and
everything
and
things
like
that,
and
that's
pretty
much
that's
pretty
much
it.
C
And
so
that's
what
the
pull
request
is
about.
There
was
some
back
and
forth
about
the
because
I
originally
had
set
it
up
with
the
title
was
a
dash
dash
wide
option,
because
in
in
my
head,
like
when
you
do
a
git
pods
dash
y
o
wide,
it
gives
you
the
node
information.
C
A
A
A
I
think
that
if,
if
the,
if
the
flag
is
called
dash
dash,
node
name,
a
lot
of
people
will
confuse
that
this
flag
is
for
not
showing
the
name
but
for
filtering
the
node
name,
and
that
will
be
my
initial
thought
overall,
the
feature
I'm
I'm
I'm
on
board.
I
definitely
like
it
and
I'm
I'm
starting
to
think
why
not
just
make
it
by
default
part
of
the
output,
it
seems
reasonable
thing
to
have
there.
A
A
Maybe
a
follow-up
would
be
to
have
the
ability
to
filter
by
a
particular
node.
If
you
know
that
you
have
a
particular
node
well,
you
can
grab
that's
true.
That's
one
of
the
options
and
even
just
having
the
the
node
name.
There
is
perfectly
valid,
even
by
default,
I'm
not
sure
how
others
are
are
looking
at
it.
F
Yeah,
I
think,
like
the
the
feature,
makes
sense
the
having
it
by
default.
I
think
like.
If
we
were
designing
this
today,
then
that
would
probably
make
sense
having
it.
G
I
mean,
while
we
officially
do
not
support
people
running
like
aux
script,
to
parse
out
data
from
these
things,
putting
like
changing
that
output
format
would
break
all
those
scripts.
So
if
we
want
to
enable
it
with
a
flag
first,
that
seems
like
a
very
low
risk
sort
of
thing
to
do,
and
then
we
can
decouple
that
from
the
decision
of
are
we
changing
the
default
output.
G
C
Yeah,
so
it
the
code
was
the
code's
pretty
simple:
that
there
is
one
extra
call
that's
made.
If
you
pass
this
flag
in
to
get
the
pods
which
could
be
could
be,
I
guess
optimize
a
little
bit
more
but
but
yeah.
C
So
if
you
did,
I
would,
I
would
suggest,
probably
not
doing
this
by
default
just
because
of
that,
because
you're
going
to
add
in
one
extra
call
initially,
I
wanted
to
have
the
api
return,
this
data
by
default,
but
when
you
do
the
top
pods
call
from
the
api,
but
that
was
like
I
was
like
okay.
This
is
a
little
bit
more
involved
than
I
really
wanted.
I
really
wanted
to
get
into
it
right
now,
so
it
was
just
like.
Let
me
just
make
an
extra
call
to
the
api
to
get
all
the
pods.
C
A
I'll
try
to
have
a
look
at
this
particular
pr
later
this
week.
A
I
think
the
only
the
overall
agreement
is
that,
yes,
it
it
should
be
there
without
any
question
the
only
the
only
question
is
whether
we
want
to
have
a
new
flag
or
whether
the
dash
of
white
could
be
retrofitted,
because,
theoretically,
that's
that's
modifying
the
output
and
that's
something
that's
a
pattern
that
we
try
to
stick
with
across
the
board.
So
if
you're
using
get
your
or
pretty
much
any
printing
command,
you
are
already
familiar
with
that.
A
If
you
pass
dash
o
y
you'll
you're
you'll
get
the
extended
output,
so
I'm
personally
leaning
towards
that
one
because,
like
I
said
initially
the
node
name,
my
initial
thought
would
be.
This
is
how
I
filter
and
not
show
node
name.
H
H
I
was
I
was
going
to
ask,
because
I
I
did
the
first,
the
first
review
into
this
pr
and
about
the
node
name
is
instead
of
the
internal
ib.
Actually
I
asked
that
because
I,
I
guess
for
most
most
of
the
users,
makes
more
sense,
the
node
name
and
not
the
internal
ip
right.
So
I
I
was
wondering
instead
of
making
this
this
second
call
when
you,
when
you
create
a
button-
and
this
is-
and
it's
already
scheduled,
the
controller
puts
the
node
name
as
a
specification.
I
think.
H
So
so
I
don't
think
you
need
to
to
make
another
call
and
this
the
pods
from
the
nodes,
but
on
the
other,
on
the
other
hand,
I
don't
know
if
you,
if
you
insert
this
like
a
manifest
inside
the
etc
kubernetes
manifests,
if
this
is
part
of
the
the
specification
that
the
api
returns
also
when
you
do
a
cube,
ctrl
top
pods.
So
I
will
just
take
a
look
into
that
because
probably
this
this
might
be
health
helpful.
This
is
part
of
the
specification
of
the
of
the
pod.
I
think.
G
A
Yeah
good
point
I'll
I'll
synch
with
ricardo
and
josh
about
the
the
shape
and
some
videos.
Okay,
let's
jump
to
the
second,
which
is
fill
with
customized
plugin
for
kit
builder.
G
I'll
just
give
you
a
quick
summary.
I
could
give
a
demo,
but
there's
a
lot
going
on,
and
so
I
don't
want
to
I'll
do
that
on
the
end,
if
there
happens
to
be
my
the
tl
dr,
is
that
two
builder
reduces
customize
for
generating
its
configuration.
G
G
It
involves
generation
of
certs
and
wiring
those
into
different
places
in
terms
like
base64
and
coding
stuff,
or
I
believe,
and
so
so
it's
not
really
just
about
taking
some
static
set
of
yemel
and
patching
or
manipulating
it.
It's
actually
about
generating
yaml.
G
G
So
I
wrote
a
little
I
converted,
essentially
all
of
the
builder
scaffolding
for
customize
into
a
customized
plug-in,
and
so
now
it
goes
and
generates
straight
from
source
as
part
of
the
customized
pipeline,
and
I
can
give
a
demo
at
the
end
if
people
want
and
if
there's
time,
okay,
otherwise
it's
just
a
fyi
if
you're
interested
in
sophisticated
cases
of
using
customized,
plugins
or
functions.
This
is,
I
think,
a
pretty
good
case.
A
Yeah
I'll
be
interested
in
seeing
demo,
let's
jump
to
to
sean
and
harsh
first
and
then,
if
we
get
the
time
we
can
return
sean.
You
want
to
talk
about
the
commands
and
headers
sure.
I
So
this
is
mostly
just
informational,
and
I
put
the
pr
for
for
implementing
this
in
in
our
docs.
This
code
is
almost
all
phil's.
He
he
actually
had
gone
quite
a
he's,
gone
down
the
path,
quite
quite
a
ways,
so
thanks
phil,
so
for
for
quarters
and
releases
now,
we've
said
we're
going
to
address
this
and
we
actually
have
now.
I
We
actually
got
the
the
kepp
into
the
new
format
and
got
the
production,
readiness
review
and
got
approved
for
this
release,
and
we
have
this
pr,
I'm
going
to
be
adding
some
more
tests.
It
actually
so
I've
manually.
I
That
this
works,
and
so
I'll
be
so,
if
you're
interested
in
this.
This
please
comment
or
reach
out
to
me
in
person
just
a
quick
reminder
of
what
this
gives
us.
So
this
will
give
us
on
the
api
server
in
the
logs,
we'll
be
able
to
tell
which
commands
are
are
connected
with
which
rest
calls,
and
so
we
can.
We
can
get
usage
statistics,
but
there's
some
advanced
cases
that
also
are
going
to
be
possible
with
this
feature,
including
adding
web
hooks
that
will
validating
web
pokes.
I
That
can
stop
particularly
dangerous
commands
by
looking
at
the
headers,
just
amongst
others.
So
please
have
a
look
it.
This
there's
one
really
interesting
little
side.
I
Note
I'll
add
before
I
give
up
the
floor,
which
is
that
this
it's
implemented
as
a
round
tripper,
and
there
was
one
command
in
particular
that
the
round
tripper
collided
with,
which
is
cuddle
proxy,
and
so
we
actually
had
to
disable
this
particular
feature
specifically
for
coupe
cuddle
proxy,
which
is
really
no
not
that
big
of
a
deal
in
in
my
opinion,
because
of
proxy,
is
just
firing
up.
It
doesn't
touch
the
api
server
initially
anyway.
A
Yeah
I'll
definitely
have
a
look
at
it
as
well.
I
remember
I
was
interested
in
when
phil
put
together
a
the
cap,
harsh
the
design
for
cube
cuddle
white.
E
They
were
doing
something
on
this
related.
I
I
I
don't
remember
exactly
so
just
wanted
to
check
if,
like
which
method
implementation,
should
we
go
over.
A
Okay,
I
remember
that
I
was
part
of
the
discussion,
at
least
from.
E
I
can
quickly
say
the
choices
about
like
yesterday.
E
Yeah,
so
the
thing
is
like
we
can
have
by
minus
minus
four
flag
multiple
times
in
the
same
command,
and
that
would
imply
that
we
have
hand
in
order
support
in
one
command
itself
where
or
would
be
inside.
The
string
of
the
four
condition
like
I
have
shown
there
in
the
example
and
giving
another
for
flag,
would
imply
the
and
condition
if
that
makes
sense,
but
the
other
choices.
E
That
we
have,
we
don't
support
or
we
don't
support
yeah.
We
don't
support
and
we
just
use
multiple
for
flex
to
indicate
that
we're
using
or
if
that
makes
sense
at
all.
E
A
Does
anyone
have
any
preference
because
I
weight
is
actually,
if
I
remember
correctly,
one
of
the
alpha
commands
or
beta
at
most,
so
we
can
do
a
little
bit
of
experimentation
as
long
as
we
maintain
backwards
compatibility,
maybe
we
could
even
play
with
different
approaches
through
providing
both
alternatives
and
gather
some
feedback
and
based
on
that,
we
we
can
make
a
decision.
A
A
A
E
Yeah,
it
was
yeah
yeah,
so
antoine
had
actually
suggested
the
other
way
around,
and
I
misinterpreted
it
and
came
up
with
this
thing
so
but
like
there
are
constant
pros
for
both
of
them.
So
I
just
wanted
to
discuss
it
out.
E
So
he
suggested
that
we
don't
support
or
the
like
the
one.
I
have
right
there,
so
he
wants
the
minus
minus
four
condition
to
be
multiple
times,
and
that
would
indicate
user
having
all
conditions.
And
if
the
user
wants
the
and
functionality
he
has
to
use
the
unix
and
themselves,
but
having
the
whole
experience
in
one
command
is
pretty
nice.
E
If
oftentimes,
you
know,
we
want
stuff
like
this
like
for
the
same
resource.
We
want
the
multiple
conditions
to
be
checked
so.
G
Comment
which
is
like
even
if
this
command
is
alphabeta,
it's
probably
one.
That's
like
more
designed
to
be
used
in
scripts
than
say,
like
top
command,
which
is
designed
to
use
for
humans,
so
like
breaking
like
any
backwards.
Compatibility
changes
in
flags
and
stuff
is
probably
going
to
break
anyone
who's
using
this.
In
that
context,
so
we
should
maybe
go
like
a
little
bit
extra
like
we
can
change
the
flag
name,
make
the
old
flag
hidden
right,
so
people
stop
using
it
market
deprecated
like
do.
A
I
that's
that's,
definitely
something
that
I'm
leaning
forward
is.
We
can
experiment
with
new
flags
and
see
where
what
the
response
is
I'll
try
to
respond
to
this
sometime
next
week
and
we
can
figure
out
where
to
go
from
there.
E
B
Sorry,
can
you
can
you
just
I
put
the
google
thread
for
the
email.
Can
you
just
bump
that
thread
with
a
summary
and
the
decision
that
needs
to
be
made
in
the
options?
Please?
I
think
I
think
this
fell
off
our
radar,
so
apologies
for
that.
E
Thanks
yeah,
neither
of
the
changes
actually
would
be
like
would
make
it
backwards
compatible,
but
putting
both
of
them
would
actually
be
tricky
because
they
both
are
conflicting
with
each
other.
F
G
So
I
think
last
meeting
we
talked
about,
maybe
it
was
in
the
triage
meeting.
Even
we
talked
about
essentially
the
methodology.
For
you
know,
accepting
features
tends
to
be
like
we
sit
in
this
meeting
and
then
we
use
and
then
say:
okay,
that
makes
sense,
maybe
and
and
it
would
be
better
if
we
had
something
that
was
like
a
little
more
accessible
and
data-driven
and
consistent
right
and
so
tanya,
and
I
spent
some
time
chatting
and
we're
thinking
of
exploring
a
program
or
what
is
it
called?
G
You
know
basically
trying
out
a
program
to
use
some
like
icon,
emoji
indicator
on
issues
right
and
then,
when
we
do
like
our
features
or
whatever
our
our
bug,
scrub
or
feature
exploration
like
we
just
filter,
and
then
we
just
communicate
to
the
community
like
broadly
that,
like
hey,
we'll
start
looking
at
future
requests
around
20
engagements
through
this
emoji
or
something
like
that.
So
like
that
way,
we
ensure
that
there's
at
least
20
individuals
who
are
excited
about
this,
which
should
be
possible.
G
Like
you
know,
everyone
has
20
friends
right,
but
I
I
still
think
that's
that's
better
than
having
you
know
no
bar
for
a
level
of
engagement
from
losers
and
then
us
just
kind
of
making
a
judgment
call
each
time.
G
So
I
I
need
to
follow
up
with
tanya
to
figure
out
where
we
are
on
that.
I
think
she
was
going
to
basically
write
up
a
a
proposal
just
like
a
very
light
proposal
of
like
here's.
What
the
emoji
is,
here's,
how
many
we
want
to
see.
G
A
A
Write
what
write
the
yeah?
Actually
someone
will
say
I
want
to
see
this
being
done.
It
gets
the
the
necessary
20
or
whatever
the
number
the
magic
number
will
be,
and
then,
on
top
of
that
the
author
or
someone
else
will
say
I
will
do
it.
G
Yes,
that
was
the
second,
so
there's
essentially
two
dates.
I
guess
gatekeeping
is
not
a
popular
term
but
checks.
I
don't
know
essentially
two
checks
that
are,
you
know,
necessary
for
us
to
make
strong
statements
about.
You
know
the
taking
on
a
feature
being
a
good
decision,
the
first
one
being
that
there
has
to
be
some
amount
of
desire
from
the
community
to
have
the
feature
and
the
second
one
being
that
it
has
to
be
it's
going
to
have
to
have
the
right
level
of
ownership
right.
G
So
maybe,
like
20
people
want
to
feature,
but
it's
like
100
people
want
a
feature,
but
it's
like
gonna
require
a
lot
of
work
and
no
one's
really
signed
up
for
it,
and
so
it
doesn't
really
matter
because
no
one's
gonna
do
it.
So,
whatever
orders
you
know
versus,
maybe
something
where
only
like,
like
the
node
name,
I
think,
is
a
perfect
example
where,
like
the
bar
for
that
should
be
relatively
low.
Given
that
it's,
you
know
not
creating
a
ton
of
work,
there's
someone
who
wants
it.
G
So
we
just
need
to
make
sure
that,
like
the
community,
like
there's
some
baseline
level
of
engagement
from
the
community,
but
not
not
consider
staffing
as
part
of
that
so
yeah.
Those
are
the.
I
agree
with
both
of
those
I
think
the
other
stuff
like
asking
like
we
could.
G
We
could
throw
in
there
like
oh
tests
and
like
this
sort
of
stuff
right,
but
I
think
we
probably
don't
want
to
do
that,
because
I
think
that
this
is
more
of
just
like
a
filtering
mechanism
for
us
to
steer
conversation,
and
so
we
should
probably
ask
for
people
to
do
the
cap
or
do
the
readme
or
do
that
stuff.
After
after
we've,
given
like
a
quick
review
on
the
feasibility.
A
Would
that
also
include
plugins,
because
I
remember
that
we
had
a
several
conversation
in
the
past,
where
we
are
backing,
basically
any
acceptance
criteria
for
for
plugins.
So
if
someone
will,
you
know
turn
to
us
with.
Oh,
this
is
my
plugin.
I
would
really
want
to
see
this
plugin
be
part
of
the
the
main
cube
cuddle,
because
it
will
be
valuable
as
a
default
with
that
proposal
also
cover
plugins,
where
someone
is
actually
interested
in.
You
know
contributing.
G
Also
has
to
contain
the
life
cycle
right
of
like
so,
you
want
to
add
a
new
command
right
start
with
a
plug-in
put
it
in
this
particular
repo.
Here's,
the
directory
put
it
in
here's
like
the
process
for
getting
it
visible
in
cue
control
as
part
of
help,
so
that
people
know
to
install
it
right:
yada,
yada,
okay,
here's
how
many
you
know
create
an
issue
get
this
many
engagements
on
it,
then
we'll
upstream
it
right
that
sort
of
thing
for
new
commands
and
then
also
for
additional
flags
right
to
existing
commands.
G
It
becomes
okay,
like
is
the
flag
hidden
by
default
or
not
is
it.
You
know
opt-in
versus
opt-out,
the.
G
A
Yeah
awesome
that
sounds
great
thanks.
A
lot
phil
for
starting
that
work.
A
Okay,
cool.
I
think
we
have
about
22
minutes.
So
that's
sufficient
time
for
both
send
ups,
and
since
we
promised
phil
a
little
bit
of
time
for
his
demo.
A
Yeah
but
I
I
was
curious
to
see
the
demo
myself
as
well
of
that.
Let
me
quickly
make
a
go
post
and
stop
sharing
the
screen
I'll
give
it
a
try.
J
G
A
I
thought
that
we
were
doing
the
demo.
That's
why
I
called
you
up.
J
Okay,
so
we've
been
working
on
getting
the
the
version
of
customize
in
cube,
cuddle,
updated
for
quite
some
time
and
there's
a
pr
out
now
that
does
the
job
and
right
now
we're
working
on
trimming
some
dependencies
so
that
we
don't
minimize.
So
we
minimize
the
number
of
dependencies
we
introduced
into
coop
cuddle,
but
the
good
news
is
is
that
a
lot
of
people
have
been
waiting
for
the
ability
to
crank
up
the
version
numbers
of
other
dependencies.
J
Things
like
yaml,
libraries
and
whatnot,
and
the
existing
version
of
customizing
cube
cuddle
was
kind
of
blocking
that
so
that's
all
being
eliminated
now,
and
I'd
really
like
to
get
this
pr
in
before
the
code
freeze.
So
we'll
trim
out
some
of
the
depths
as
much
as
we
can
and
hopefully
we'll
get
that
in
before
everything
freezes
up.
I
Sorry
to
interrupt
so
I
saw
the
on
the
pr.
There
was
some
talk
about
removing
starlink
dependency.
J
B
Hey
jeff
did
is
this:
are
we
going
to
have
the
same
issue
in
like
you
know,
six
months,
12
months,
when
we
want
to
update
the
version
again
or
it's
we're
good?
No,
no.
This
is
a
fundamentally
different
approach.
J
Nice
awesome,
it's
just
awesome.
There
was
a
over
the
past
year.
You
know
the
switches
past
two
years
now
I
guess,
but
the
big
switch
over
to
go
modules,
which
are
wonderful,
eliminated,
a
lot
of
things,
a
lot
of
the
hacks
that
were
done
to
get
customized
in
there
in
the
first
place.
So
now
it's
just
conventional
imports.
I
J
Yeah
yeah,
it's
all
depending
on
well-semvered,
or
at
least
I
think,
they're
well
sunburned
modules.
At
this
point.
G
Are
the
does
this
give
us
that
comment
and
structure
to
keep
keeping
that
stuff,
or
is
that
not.
J
Oh,
so
there's
I'm
not
saying
that
so
there's
cleanup
work
that
can
now
be
done.
There's
this
constructured
stuff,
there's
interfaces
that
are
still
existing
in
in
customized,
which
allowed
the
ability
to
have
you
know
inject
either
the
api
machinery
dependencies
or
the
kml
stuff.
J
The
there's
a
lot
of
shouldn't
say
a
lot,
but
there
are
layers
in
customized
that
can
now
be
removed,
but
the
actual
depths
are
gone.
So
there's
a
there's,
an
issue
in
customized
to
do
the
cleanup
work,
but
that
cleanup
work
does
not
block
in
any
way
this
reintegration.
B
B
J
Yeah,
I
can,
I
can
write
the
blog
post
and
try
to
identify
what
features
break
so
anybody
who's
currently
using
customize
standalone
customize.
They
can
switch
over
to
using
coop
cuddle
customize
without
any
pain,
because
it's
the
same
thing
modulo
the
hashicorp
dependencies
that
we've
we've
cut.
J
J
J
B
A
Maybe
even
instead
of
because
I'm
not
a
fan
of
automatically
fix
your
files,
maybe
it
would
be
even
better
to
be
explicit,
because
if,
if
we
are
on
the
fly,
fixing
something
for
people,
it
won't
be
fixed
permanently
on
the
disk.
So
it
would
be
better,
like
eddie
said,
put
a
warning
put
a
suggestion:
how
to
fix
it
by
running
the
command,
you'll,
get
a
permanent
fix
in
place
and
then
just
rerun.
The
command.
B
If,
if
do
so,
I
realized
that
we're
getting
close
to
code
freeze-
and
I
definitely
don't
want
this
to
miss
because
of
this
stuff,
if
jeff,
if
you
scope
this
out
and
you
put
out
some
calls
for
help
like
I'm
happy
to
jump
in
and
figure
out
what
I
can
do
and
I'm
sure
other
people
are
as
well.
So
don't
don't
feel
like
we're
just
throwing
this
at
you
last
minute
here,
yeah,
okay,.
G
J
A
That,
okay
cool,
that
that
was
a
major
item
to
take
thanks
a
lot
jeff.
I
know
you've
spent
numerous
hours
days
weeks,
yeah
even
months,
I
would
say
cranking
this
up
and.
H
Yeah,
I'm
sorry
about
putting
this
at
this
meeting
agenda,
but
I
am,
I
will
not
be
able
to
attend
the
script
next
one.
This
is
a
an
issue
about
adding
endpoints
in
endpoints
count
in
in
cuba.
Ctl
get
services,
I
guess,
and
I
I
thought
it
was
going
to
be
an
easy
one.
I
I
think
eddie
also
thought
it
was
going
to
be
a
an
easy
one,
but
yeah
it
happens
that
the
way
that
api
machinery
does
things,
I
don't
think
we
can,
we
can
cross
between
objects.
H
E
A
H
G
A
Would
have
to
look
deeply,
but
I
know
that
in
some
get
for
a
particular
resource,
we
are
reaching
for
additional
elements,
but
I'm
not
100
sure
I'll,
dig
through
I'll
dig
through
the
code
and
I'll
leave
an
answer
in
that
in
in
that
issue,
whether
it's
feasible
or
we
will
just.
We
will
push
it
towards
this
cry.
Brother.
H
I
appreciate
that
because
I
spent
something
like
four
hours,
trying
to
figure
out
what
was
printing
then
and
then
figure
out.
That
was
the
api
and
what
kids.
A
Ricardo,
hang
me
on
slack
I'll
point
you
to
where
the
table
printers
are
which
are
responsible
for
the
output,
and
this
is
something
that
other
art
others
might
be
interested
I'll
I'll.
Definitely
I
can
point
folks
it's
basically
package
internal
printers,
something
like
that.
Yes,.
F
G
We
kind
of
posted
those
on
yes
issues.
A
Nick
or
paul
do
you
have
anything
from
the
koi
front?
I
know
that
there's
there's
been
some
movement
and
and
moving
koi
to
to
cube6.
K
K
I
think
we
have
like
two
or
three
stragglers
and
so
we're
forcing
some
that
we
know,
and
then
the
ones
we
don't
know
we
need
to
figure
out.
So
that's
yeah,
this
kind
of
stuff.
The
long
tail
takes
forever
to
get
these
little
things
squared
away.
I
did
have
a
question.
This
is
sort
of
a
generic
question,
but
is
there?
Is
there
a
good
way
for
us
to
distinguish
system
namespaces?
A
K
A
K
Okay,
it'll
be
great
if
there
is
a
label
yeah.
So
in
short
of
that,
in
the
release
we
just
made
it
release
10.2
hook
and
fire
preparation
for
this
repo
transfer.
That
introduces
some
of
these
heuristics
just
to
try
to
do
a
better
distinction.
K
So
now,
if
you
open
up
the
namespace,
switching
widget
you'll
be
able
to
type
that
find
you'll,
be
able
to
see
a
distinction
between
our
characteristics
for
distinctions
and
if
other
providers,
like
microsoft
or
google
or
aws,
want
to
come
to
the
table
and
add
additional
heuristics,
we
made
it
a
very
clear
way
for
adding
provider
level
heuristics
cool
other
than
that.
I
want
to
see
the
demo
yeah
me
too.
So
I
relinquish
my
time.
G
Great
okay,
so
what
I
got
here
is
there's
this
new
command
I'll
roll
through
some
examples
first
and
then,
and
then
then
from
there
you
can
then
I'll
show
you
kind
of
how
it's
implemented.
So
the
first
example
is
like
doing
this
as
a
standalone
command.
I
think,
there's
you
know
reasons
that
this
makes
sense.
G
The
user
might
want
to
do
this
right,
and
so
you
do
something
like
this
and-
and
you
can
see
here,
the
resource
file
right,
it
says
to
builder
project
and
then
it
has
a
project
directory.
So
this
is
where
the
builder
code
is
and
it
has
an
image,
and
so,
if
I
you
know,
run
this
thing,
I
didn't
test
this
beforehand
by
the
way.
D
G
G
By
looking
at
the
types,
if
I
want
to
do,
this
is
using
the
function
framework
as
part
of
that's
part
of
customize.
I
can
give
it
patches
and
so
then
it
will
apply
the
patches.
So
in
this
case
it's
like
replicas
five
set
and
then
without
the
patch,
it's
replicas
one
here,
so
you
can
so
that's
kind
of
like
the
standalone
mode.
I
think
it's
kind
of
an
interesting
pattern
for
if
you
have
like
istio
or
something
and
you
want
to
install,
but
it
also
is
done
as
a
customize.
G
This
is
where
I
think
it's
cool.
It
shows
like
the
flexibility
of
this
approach.
So
what
I
have
here
is
I
took
you
know
that
alpha
config
gin
command,
which
I
I
just
ran,
and
then
I
go
and
install
it
as
a
customized
plug-in.
So
this
is
just
the
the
standard
plug-in
approach.
Right
has
this
little
script
here
and
then
I
have
this
customization
ammo.
G
So
I
just
give
it
as
a
transformer,
and
this
is
that
same
config
you
saw
before,
and
so,
if
I
run
customize
build
on
this,
you
can
see
that
it
generates
the
same.
The
same
thing
from
from
the
customized
piece,
and
then
I
have
some
test
data
here.
I
think
we'll
show
you
kind
of
like
where
this
gets
interesting
is.
G
I
can
do
something
like
say:
okay,
enable
web
hooks
just
as
a
field
instead
of
commenting
out
customized
stuff,
because
there's
now
there's
an
api
right,
and
so
that
will
like
trigger
whether
web
hooks
get.
You
know
generated
there's
one
here,
which
is:
where
is
it?
G
Actually?
I
may
not.
I
mean
I
have
an
example
of
a
generating
insert,
but
it
will
actually
generate
a
certain
wired
in
for
development
for
web
hooks
so
that
you
don't
have
to
like
spend
so
much
time
doing
that
which
was
kind
of
a
pain
it
can
work
with
component
config
right,
so
component
config
gets
wired
in
slightly
different.
G
So
if
you
give
it
like
the
path
to
a
component
config
file,
it
will
use
like
that
methodology
for
generating
the
configuration
and
so
kind
of
the
nice
aspects
of
using
customize
is
you?
Can
you
can
use
all
the
other
customized
stuff?
So
you
can
patches
and
resources
and
use
this
as
product
faces,
and
do
you
know
config
map
and
secret
generation
or
stitch
together
a
couple.
G
You
know
customize
files
from
this,
so
so
it
works
just
like
any
other
customized
transformer,
but
it's
replaces
just
massive
amounts
of
configuration,
like
the
alternative
to
this,
for
instance,
is
if
you
look
at
like
a
project
here.
G
This
is
you
can
see
this
directory.
Is
everyone
able
to
see
this
directory
on
the
side
yeah?
We
can
see
it
so
you
have
like
you
know
these.
This
is
like,
instead
of
just
to
turn
off
prometheus
right
like
how
do
I
turn
on
or
off
prometheus?
I
gotta
go
track
down
like.
Where
is
the
prometheus
thing
right
so
over
here
and
then,
like
you
know,
this
is
the
approach
here
right
so
I'd
to
to.
G
I
turn
off
prometheus
here
and
then
and
then
for
me,
this
someplace
else,
where
I
turn
off
web
hooks
here
and
so
you're
kind
of
like
searching
around
for
for
these
comments
to
figure
out
how
to
stitch
these
things
together.
So
it's
just
a
matte,
it's
a
way
of
introducing
abstractions
to
as
like
kind
of
a
first
class
notion
to
customize,
and
it
allows
for
two
builder.
This
is
useful,
because
one
thing
this
allows
us
to
do
is
extend
that
abstraction.
G
So
in
this
case,
if
you
want
to
have
a
specialized
two
builder
project
that
generates
configuration
differently
like
for
operator
sdk,
for
instance,
you
just
layer
in
your
operator
sdk
and
it
will
transform
the
output
of
code
builder
as
as
needed,
and
so
here's
like
an
example
of
like
a
framework
and
there's
kind
of
two
methods
we
provide
and
they
can
be
used
together
independently.
G
The
first
is
sort
of
like
allowing
you
to
codify
customize
pat,
like
what
you
do
with
customized
patches
and
stuff,
as
kind
of
these
templatized
patches
right.
So
so,
in
this
case,
this
is
the
whole
extension
that
does
the
replicas
patch,
and
so
what
it
does
it
says.
Okay,
I
want
a
new
template
command.
G
Here's
the
api
api
structure
has
this
replicas
up
here:
here's
some
patch
templates,
there's
a
list
of
them,
and
so
it
says:
okay,
here's
like
the
patch
that
you
generate,
and
then
you
apply
it
to
a
deployment
named
controller
manager
right.
So
this
is
essentially
generating
a
customized
patch
programmatically.
G
You
can
generate
resources
using
templates
as
well.
The
other
way
of
doing
this
is
like
a
more
sophisticated
approach.
This
is
like
the
more
kml
where
maybe
you
don't
want
to
patch,
for
whatever
reason
you
want
to
manipulate
fails
directly,
you
want
to
inject
certain
things,
and
so
this
is
an
extension
that
does
the
same
thing
but
is
written
differently
where
it's,
this
template
command.
It
gets
a
resource
list
and
then
it
selects
gets
matches
of
the
deployment
controller
manager
manually.
E
G
G
Filter
these
are
all
done
as
filters
under
the
covers
see.
So,
if
you're
familiar
with
that
approach,
where
this
framework
takes
different
filters
and
then
you
know
patches,
directories
of
patches
and
directories
of
resources-
and
so
you
can
say,
okay
apply
these
filters,
so
controller
gen,
filter
and
assert
filter.
G
The
cert
generation
filter
generates
a
certificate
using
cf
ssl
and
then
injects
it
in
the
right
places,
so
it
like
gets
and
finds
it
finds
matches
for
validating
and
mutating
web
hooks,
for
instance,
so
it
generates
assert,
finds
different
web
configurations,
looks
for
the
web
field
and
then
goes
and
creates,
like
the
ca
bundle
for
instance,
or
or
wires
in
the
service
and
same
the
custom
resource
definition
for
conversion
web
hooks.
It.
Someone
does
something
similar
here,
so
it's
able
to
like
manipulate
those
resources.
G
This
is
that
more
sophisticated
approach
of
like
setting
the
different
fields
that
it's
observing
then
then
there's
simpler
ones,
which
is
like
so
that's
like
at
the
far
end
of
you
know
more
complicated
and
then
on
the
side
of
the
simpler
right.
You
can
actually
just
create
these
directories
of
templates.
G
This
looks
you
know
this
introduces
the
capabilities
of
like
helm,
charts
to
some
extent
right,
like
the
the
methodology
that
you
know
the
helm
charts
would
use
it
kind
of
allows
you
to
write
them
in
a
similar
way,
and
so
you
can
say,
for
instance,
if
this
particular
thing
is
enabled
this
is
derived
from
the
api
type.
So
this
is
a
function
on
on
the
api
call.
If,
if
the
api
says,
controller
managed
enabled
then
go
generate
this
resource
and
then.
E
G
Know
add
it
to
the
set,
and
so
so
in
this
case
the
pattern
is
just
create
a
bunch
of
files
and
then
check
on
the
api.
If,
if
those
particular
resources
are
able
and
generate
them,
patches
are
similar,
you
can,
you
can
generate
patches
the
same
way
and
then,
if
you
look
back
at
the
command
the
way
these
are
wired.
E
G
You
just
say:
okay,
take
all
the
templates
from
this
directory.
Take
the
patches
from
these
directories.
Patch
is
a
little
more
tricky
because
they
have
selectors
right,
so
you
only
patch
certain
resources,
those
a
little
more
complicated
than
just
directory
and
then
and
then
you
can
do
post-process
filters.
So
in
this
case
we,
for
instance,
sort
the
resources
and
and
do
that
sort
of
stuff,
and
then
the
the
templates
are
actually
packaged
up
with
the
command
using
packager.
G
So
this
means,
while
they're
defined
as
file
templates
down
here
there
they're
compiled
into
the
binary
with
go
gen
generate,
and
so
this
command
is
the
command
that
can
be
run
either.
What
the
framework
does
is
it
enables
you
to
run
it
either
as
a
plug-in
approach.
Right,
it'll
accept
that
input
and
it
will
look
for
it
or,
as
a
you
know,
a
function.
G
It
accepts
both
the
customized
plug,
plug-in
input
or
the
other
newer
style
with
the
resource
list,
or
it
can
also
be
run
as
a
standalone,
and
this
framework
basically
dynamically
filters
off
of
that,
and
it
allows
you
to
do
pretty
sophisticated
stuff
within
the
within
the
templates
themselves,
because
you're
you're,
defining
these
structs
as
basically
api
types
right,
and
so
you
can.
This
is
this
function
right,
allows
you
to
check
whether
components
are
enabled
and
invoke
it
from
the
the
go
template
the.
G
The
way
is
not
a
fixed
requirement,
it's
just
an
ex.
You
know
a
simple
way
of
writing
resources.
It's
not
that
the
customize
itself
cares
nothing
about
go
templates,
it
doesn't
know
they
exist
at
all.
It's
just
a
way
that
the
framework
enables
users
to
create
a
resource
configuration.
That's
one
of
its
options.
L
I
just
want
to
comment
that
I've
been
working
on
iteration
on
the
framework,
since
we
just
saw
some
of
the
the
code
there.
I
have
an
open
pr
on
customization
draft,
probably
going
to
turn
it
into
a
real
pr
today
that
takes
some
of
those
interfaces
that
phil
was
just
showing
and
makes
them
even
more
powerful
by
doing
things
like
decoupling,
the
the
cobra
aspects
from
the
not
cobra
aspects,
centering
it
even
more
on
filters
and
making
it
multi-version
multiversion
api
compatible.
J
A
Yeah
that
looks
interesting.
Thank
you
very
much
all
we're
five
minutes
over
top
of
the
hour.
So
sorry
for
that,
but
thank
you
very
much.
All
it
was
nice
talking
to
you
as
usual
and
see
you
next
week
for
the
block
blocker
box,
grub
for
remember
correctly,
bye,
y'all,.