►
From YouTube: Kubernetes SIG CLI 20210818 - bug scrub
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
Good
morning,
good
evening,
good
afternoon,
depending
on
where
you
are
today
is
august
18th-
and
this
is
another
of
our
monthly
box
crops-
our
host
will
be
eddie,
so
take
it
away
from
here,
eddie.
B
Otherwise
I
had
one
that
we
should
look
at,
so
my
co-worker
actually
ran
into
this
the
other
day.
If
you
have
a
crd
with
a
top
level
directive
of
x,
kubernetes
preserve
unknown
fields.
True,
it
breaks
cube
control.
Explain
so
here
was
an
example
of
it
commented
out
and
we
have
all
of
our
descriptions
and
whatever
and
then
with
it.
B
So
I
tagged
the
api
machinery
and
it
looks
like
they
got
back
yesterday
that
this
appears
to
be
a
cube
control
problem.
The
field
is
valid
on
a
crd,
so
I
didn't
know
if
the
field
was
valid
at
a
top
level
and
according
to
feta,
it
is
so.
A
Yeah,
if
I
remember
correctly,
open
api
is
not
published
when
the
kubernetes
preserve
phantom
fields
is
set.
That
was
part
of
the
the
structural
schema
or
something
along
those
lines.
I
can't
remember
specifically
what
was
the
name,
so
it's
possible
that
what
you're
seeing
is
basically
what
the
open
api
has
and
if
they
open
api
smp,
you
just
basically
got
an
empty
information,
so
I
wouldn't
call
it
as
an
error.
B
A
I
would
it
would
probably
omit
only
that
single
field,
because
that
field
does
not
follow
the
schema
and
I
think
it's
being
omitted
in
the
open
api.
It.
B
C
So
something
that
may
or
may
not
be
relevant
is
that
there's
been
a
code
change
specifically
for
x,
kubernetes,
preserve
unknown
fields
that
antoine
police
did
within
the
last
two
to
three
months
there
there
has
been
changes
in
the
code,
so
it
would
be
interesting
to
to
see
if
this
actually
is
previous
versions.
Also.
B
B
As
for
our
triage
party,
I
do
have
infrastructure
provisioned.
We
have
a
google
group
that
was
created
and
I
believe
katrina,
myself
and
mache
were
added
to
this.
I
think
anyone
can
add
themselves
with
a
pr,
but
this
lets
us
we.
Basically
we
have
our
own
namespace
in
the
aaa
cluster
for
kubernetes
org
and
we
can
deploy
triage
party
there.
So
I
just
have
to
copy
paste
the
manifest
and
do
the
little
shell
script
and
we
should
be
good,
so
that
is
on
my
list
for
this
week.
B
Oh
okay!
Yes,
this
is,
we
knew
this
was
a
problem
with
the
go
completion.
You
can't
pass
it
together
to
do
completion.
I
think
this
is
fixed
on
the
newest
version
of
cobra,
which
is
also
on.
I
have
a
pr
open
that
is
upgrading
that
so
this
one's
accepted,
the
scribe
pod,
doesn't
take
into
account
the
new
fields
of
events.
A
Yeah,
that's
that
should
be
pretty
simple
fix.
We
can
probably
mark
this
one
as
a
good
first
time
issue.
A
We,
the
events,
have
a
pretty
basically
hard-coded.
How
the
event
looks.
Event
should
be
printed,
so
it
should
be
pretty
straightforward
for
anyone
to
figure
out
and.
B
Update,
okay,
so
we'll
say,
trio
is
accepted,
help
wanted.
A
B
B
A
B
You
should
look
inside
these
files
and
this
is
most
likely
where
it
is.
So
I'm
happy
to
take
a
look
at
and
type
that
up
to
see
if
we
can
get
someone.
A
Probably
there's
a
single
file
that
you
just
need
to
look,
there's
probably
something
that
is
called
print
event
and
something
and
it
lives
in
describer
that
go
or
something
like
that.
B
Okay
support
multiple
okay
katrina
and
lao
are
working
on
that
one.
Any
thing
we
need
to
talk
about
for
that
and
allows
the
pr
open.
E
I
lgtm
yesterday
and
antoine
was
taking
a
look
this
morning.
We
had
some
questions,
we
answered
them.
I
think
it's
pretty
much
ready
to
go.
A
Yeah,
that
sounds
reasonable.
That's
probably
a
an
error
in
the
how
we
are
parsing
the
file
does
it
that
it
doesn't
deal
with
pointers.
As
the
author
is
pointing
out.
B
A
B
I
to
take
a
look
if
these
types
are
pulled
in
by
config,
as
they
start.
A
B
B
A
And
all
in
all,
I
think
I
agree
with
what
doug
just
wrote
here.
Given
our
past
experience,
how
we
try
to
move
the
config
file
changing
the
name
will
be
similarly
problematic.
So.
E
A
C
There's
a
relatively
special
set
of
exit
codes
for
diff
that
it's
actually
explained
in
the
in
the
help
for
ku
kettle
diff.
So
it's
not
just
zero
and
one
like
many
other
commands.
C
Follow
but
there's
zero
one
and
two
and
now
there's
also
server
side
diff
and
has
been
for
quite
a
while.
A
Right
but
server
side
diff
will
probably
we
won't
even
get
to
that
one,
because
the
the
local
one
will
fail
when
trying
to
read
a
file
that
does
not
exist.
So
probably
I
would
I'll
be
curious
to
see
whatever
regular
there.
E
It
can't
be
reproduced
there
a
it
makes
sense
to
me
that
it
shouldn't
exit
zero
if
the
file
doesn't
exist
because
customization's
failing,
but
if
it
can't
be
reproduced.
That's
another
problem.
A
Definitely
it
should
not
exist
exit
with
a
zero.
It
should
return
an
an
error
code
too
yeah,
that's
difficult,
be
triage,
accepted.
A
The
next
one
cube
cuddle
off
can
I
miss
documentation
for
ask
parameter.
A
A
I
think
we
are
explaining
and
help
that
use
cubecut
options
if
you
want
to
see
global
command
line,
options
that
apply
to
all
commands,
so
I'm
not
sure
if
we
need
any
particular
changes.
I'll
have
a
look
at
this
pr,
but
I
would
but
I'm
inclined
to
close
it,
because
you
have
all
the
information
in
the
help
command
of
any
part
of
any
command.
Basically
I'll
have
a
look
at
it.
After
the
call
and
comment
on
that,
one:
okay.
A
Okay,
I
see
that
ryan
already
has
a
pr.
It's
just
been
open
for
quite
a
while
I'll
have
a
look
at
it
after
the
call.
Okay,
okay
did
we
have
that
it
doesn't
trick?
A
Yeah,
it
looks
like
it's
not
a
cute
cuddle
particular
issue,
but
rather
the
underlying
container
system
container
engine.
A
A
I
I
don't
expect
us
to
fix
anything
so
since
this
this
does
exist
against
that
cupcake
repo
I'll
most
likely
close
it,
especially
that
yeah
it's
being
easily
reproduced
with
docker.
I
bet
that
if
I
try
this
with
cryo,
the
effect
will
be
basically
similar.
The
problem
is
that
when
you
kill
the
process,
you're
you're
only
killing
the
the
cubical
exec
and
nothing
is
or
can
actually
affect
the
original.
A
D
A
A
A
A
A
A
A
A
A
E
A
You
can
even
create
your
own
crd,
I
think
that's
called
the
field
is
called
alias
or
something
along
those
lines
and
add
whatever
you
care
about
under
this
particular
alias.
So
if
I
create
my
foo,
it
can
be
easily
added
to
all.
So,
even
if
we
deprecate
it
and
remove
it,
then
that
that's
first
of
all,
that's
a
decision
to
be
done
at
the
api
level
and,
secondly,
it
can
be
easily
recreated
by
anyone,
although
it
won't
include
the
built-ins
in
the
in
that
case,.
B
E
A
E
Right,
so
why
do
we
not
want
to
maintain
the
alias
just
too
much
work
to
get
all
meaning
actually
all
to
keep
that
list
up
to
date
or.
A
It
all
might
mean
because
you
cannot
change
the
built-ins,
but
the
problem
is
all
can
mean
different
things
at
different
clusters
and
we
will
be
getting
and
answering
issues
that
oh
I'm
seeing
this,
but
I'm
not
seeing
something
else
and
different
people
might
get
different
outputs
and
we
said
that
we
don't
wanna
do
anything
like
that.
Basically,
all
was
a
mistake.
It
should
have
been.
I
don't
know
core
resources
or
something
more
descriptive.
A
A
Yeah
you're
right
categories,
that's
that's
the
field
name.
Katrina
just
checked.
B
We
have
a
lot
of
these
everywhere.
I've
closed,
so
many
of
these
are
they
all
linked
to
here.
No,
this
is
one
of
those
things
that
I
feel
like.
We
should
do
something
about.
What
can
we
this
get
all
plug-in?
Have
we
can
we
upstream?
That,
like,
is
that
something
we'd
want
to
do
because
that's
been
our
recommended
solution.
Instead,
it's
like
hey
use.
This
get
all
plug-in.
B
B
A
I'm
not
sure
I'm
positive,
we
don't
want
to
do
it
in
the
first
place,
because
on
top
of
that,
the
different
cluster
will
have
different
resources
comes
an
issue
of
different
users
will
see
different
values
from
like
at
all,
because
it
will
depend
on
what
you
actually
have
access
to.
A
So
my
proposal
will
be
to
get
rid
of
all
entirely
because
it
will
be
simpler.
It
would
break
people.
Yes,
it
will
be
the
first
time
that
we
are
breaking
people
and,
secondly,
it
would
have
to
go
through
sick
architecture,
I
think-
or
at
least
sig
api
machinery,
because
they
own
the
api,
because
we
will
need
to
change
the
the
built-ins
specifically
to
get
rid
of
the
get
all
yes
entirely.
A
A
A
B
A
E
Personally,
I've
only
used
it
or
like
reached
for
it
in
development
clusters
with
basically
nothing
there
to
remind
myself
what
I've
like
what
junk
I've
thrown
in
it.
You
know
on
my
local
machine,
not
any
real
time
in
production
because
yeah,
it's
not
it's
not
comprehensive
enough.
A
I
can't
imagine
invoking
get
all
on
a
production,
especially
when
you're
close
to
some
some
admin
and
then
add
the
uppercase
a
to
it
and
you
will
ddos
your
cluster.
B
I
am
going
to
send
an
email
to
dev
and
ask
for
thoughts
and
feedback
there
yeah,
because
I
I
just
hate
these
things
that
we
get
in
where
it's
just
like.
We
can't
do
anything,
it's
stuck
the
way
it
is,
and
we
don't
want
to
do
anything.
You
know
we
don't
want
to
make
it
better.
We
don't
like
it
like.
I
agree
with
you
saying
that
it
was
probably
a
mistake.
It
made
sense
at
the
time,
but,
like
I.
A
I
can
admit
that
I've
been
using
ghetto,
but
that
was
like
six
years
ago,
but
I
stopped
very
very
long
time.
B
B
A
Oh
okay,
okay,
I
think
I
know
what
the
problem
is
so.
A
A
A
The
question
is,
though,
because
there
is,
there
are
two
fields
there
is
one
which
is
called
service,
account
name
and
there's
another
one
which
is
called
service
account.
I
think
they
changed
changes
recently,
but
for
backwards
compatibility,
they're,
keeping
both,
I
think
one
is
deprecated
and
the
other
one
is
not.
A
A
A
Oh
she's
on
120,
so
he's
pretty
up
to
date.
B
Okay,
we
should
do
some,
so
all
right
so
expected
behavior
is
either
it
warns
you
that
it
can't
be
updated,
or
does
that
so
does
changing
it
actually
update.
It,
though,
does
removing
it
change
it
to
default.
A
A
Because,
currently,
if
you
look
at
it,
different
tools
are
doing
it
differently.
I
know
that
gcloud
creates
an
appropriate
cubeconfig
to
use
with
your
gke
cluster
aws.
It
is
doing
this.
Similarly,
openshift
has
a
built-in
login
command
that
handles
the
creation
of
the
of
the
file.
I
guess
a
particular
cluster
kind
handles
that
one
for
you
as
well
by
writing
the
file
that
give
copy
for
you.
B
A
Yeah,
I,
like
david's
suggestion
about
having
some
kind
of
a
sum
or
aggregate,
because
you
currently
keep
cuddle
pod.
A
cube,
cuddle
top
pod
has
a
selector
and
you
can
filter
a
particular
deployment
and
david
suggested
to
have
a
sum
or
aggregate
basically
there,
because
currently
it
lists
all
the
resources
and
prints
information.
So
yeah,
that's
definitely
something
helpful
on
it.
F
C
Okay,
so
so
I
put
one
in
the
chat,
can
we
have
a
quick
look
at
that?
It
was
actually
in
the
triage
party,
and
I
I
accepted
it,
and
I
also
threw
it
or
I
signed
it
to
myself
and
threw
a
triage
accepted
on
it,
but
I
should
have
waited
before
doing
that
and
it
was
I
I
assigned
it
to
myself,
because
I've
just
been
working
in
here
with
the
headers,
but
I'd
like
your.
C
C
C
My
worry
was
that
I
had
changed
something
with
the
coupe
cuddle
headers
that
had
triggered
this,
but
it
doesn't
sound
like
it
and
that
I
I
yeah
hoping
to
get
some
thoughts
from
the
rest
of
you
guys
on
this.
One.
A
Now
that
that
looks
like
you're
just
missing
an
interface
being
implemented
by
the
command
header
round,
tripper.
A
Similar
to
what
was
pointed
out
in
in
the
other
issue
fixed
by
the
cigar,
so
yeah,
that's
definitely
something
that
you
did
introduce
sean
okay.
I
can
tell
you
that
quickly
and
yeah,
the
fix
should
be
pretty
simple,
just
implement
the
console
request,
it's
totally
fine
to
even
have
an
empty
method.
A
As
long
as
it
just
follows,
the
the
interface
that
is
apparently
expected
sure
how
that
was
triggered
would
be
nice
to
figure
out
when
or
how
this
was
triggered.
C
A
A
A
B
C
I'll
try
to
figure
out
what
the
reproduction
case
is
and,
and
try
to
you
know,
have
some
kind
of
tests
that
that
figures
that
out,
but
the
reason
that
you
know
it
got
in
was
that
it
sounds
like
it's
difficult
to
actually
get
into
this.
B
C
I'll
I'll,
I
should
have
a
pr
relatively
quickly
mache.
Could
I
send
that
to
you.
A
Yeah
ping
me
on
slack
with
that
one.