►
From YouTube: Kubernetes SIG CLI 20220525 - 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
on
where
you
are
today
is
may
25th-
and
this
is
another
instance
of
our
monthly
box
crop
calls.
My
name
is
mate
and
I'll,
be
your
host
replacing
eddie
who's
on
his
journey
back
from
from
spain
from
kubecon
last
week,
so
bugscrap,
let's
start
from
the
top
sending
over
to
the
chat,
and
let's
say:
let's
give
everyone
a
bit
to
read
it
through.
B
Seem
seems
like
a
completely
different
functionality
than
than
what
we
normally
do
for
for
create.
A
When
I
initially
wrote
that
kind
of
functionality
for
create
to
be
able
to
create
one-offs
from
a
particular
current
job,
we
never
expected
the
ability
to
to
change
the
definition
of
the
cron
job.
In
any
way,
the
only
capability
was
to
be
able
to
trigger
a
cron
job,
as
is
without
any
modifications,
and
even
that
I
was
a
little
bit
hesitant
to
add,
but
eventually
god
got
convinced
to
push
this
through.
A
I'm
not
sure
about
the
the
convert
part
how
feasible
that
would
be,
but
I
really
like
the
last
suggestion
from
the
original
author,
where
he
proposes
that
if
this
is
a
known
limitation
of
apply,
which
it
is
at
this
point
in
time,
we
could
put
a
warning
explaining
that
we're
trying
to
do
a
three-way
merge
on
resources
with
different
api
versions.
A
Obviously,
this
will
only
apply
to
the
client
side
apply,
because
the
server
side
will
not
be
affected
with
this
problem.
Also,
it
does
not
invoke
3a
merge
the
server
side
apply,
so
a
simple
fix
is
adding
the
the
warning.
If
the
versions
don't
match
anyone
interested
in
picking
this
one
up.
B
So
yeah,
so
just
so
just
a
little
background
on
the
last
applied
configuration
as
as
you
approach
that
problem.
B
So
it's
one
of
the
reasons
that
we
have
to
have
a
three-way
merge
is
that
it's
impossible
to
prune
fields
with
apply
client-side
apply
unless,
because
of
defaulting,
it's
impossible
to
know
when
fields
have
been
removed,
and
so
that's
that's
one
of
the
reasons
why
we
have
this
last
applied
configuration
it's
it's
to
give
us
the
intent
of
what
they
had
and
what
which
fields
they
they
care
about
and
and
that's
kind
of
generalized
in
the
server
side
apply,
but
but
that
I'm
just
going
over
this
stuff
in
my
head
to
try
to
actually
figure
out
what
what
would
or
what.
B
What
I
think
should
happen
with
a
with
a
conversion
during
the
apply
with
this
last
applied
configuration.
So
it's
the
the
object.
That's
there
as
well
as
the
the
last
applied
configuration
from
that
we
can
determine
which
particular
fields
the
user
cares
about,
and
whether
or
not
they've
actually
removed
some
fields,
and
so
we
could
calculate
the
patch
from
that
and
then
that
patch
then
goes
is
applied.
B
A
B
A
A
Will
work,
you
will
get
different
values,
but
if
you
have
a
consistency
between
the
last
applied
configuration
and
the
file
on
disk,
then
most
likely
requesting
the
same
api
version
from
the
server
is
feasible.
So
in
that
case,
instead
of
requesting
whatever
cron
job
latest
is
you
would
request
a
specific
version
that
matches
both
of
these,
which,
up
to
a
point
where,
where
v1
beta
1
for
project
is
served,
will
work
for
you,
of
course.
A
Eventually
it
will
stop
working
for
you,
so
you'll
have
to
deal
with
with
fallouts
from
that
and
that's
something
that
we
won't
be
able
to
address
and
most
likely
the
apply
code
would
have
to
account
for
the
fact
that,
oh
I'm
requesting,
let's
say
the
v1
beta
1,
but
it's
not
served
by
the
client
by
the
server
anymore.
A
In
which
case
you
would
fall
out
you,
you
would
fall
back
to
the
current
behavior
and
try
doing
the
best
effort
with
a
warning
that
oh
there
is
a
mismatch,
I'm
trying
to
figure
out
what
would
happen
and
how
to
best
approach
the
situation
when
I
have
a
last
supply
configuration
from
before
the
upgrade.
A
A
That's
where
the
situation
is
a
little
bit
more
tricky,
because,
yes,
you
can
request
from
the
server
older
version
like
we
just
talked,
but
converting
the
the
file
on
the
disk
might
cause
some
troubles.
A
So
that's
the
two
different
use
cases
that
I
can
think
of
and
that
you
would
have
to
look
when
you
will
be
going
through
this.
That's
why
the
the
simplest
approach
that
I
can
think
of
right
out
of
the
bat
is
what
the
author
proposed
is
just
documenting
this
with
a
warning
that
oh
there's
a
mismatch
so
and
the
simplest
approach.
A
The
simplest
solution
is
to
update
the
api
version
and
force
the
apply
if
something
fails
on
it
or,
in
the
worst
case,
a
a
full
replace
on
the
resource.
A
My
voice
just
does
not
want
to
cooperate.
I
think
we
talked
about
it
last
time
and
we
requested
additional
information
and
apparently
it
looks
like
we
didn't
get
any.
A
Okay,
so
his
his
answer
basically
aligns
with
the
field
is
not
set.
The
question
is:
can
we
verify
that
the
field
actually
exists
on
the
resource
it's
just
empty?
A
If
so,
then
we
should
most
likely
be
able
to
to
wait
on
it,
and
something
tells
me
that
we
should
get
that
information
if
a
field
is
a
proper
value.
D
A
Yeah,
that's
that's
what
I
recall
as
well,
because
yeah
mark
you
wanna,
you
wanna
pick
it
up
yeah!
I
could
look
at
it.
Okay,
what's
your
github
handle.
D
That's
just
my
m
pocket.
D
I'll
I'll
just
I'll
assign
it
myself,
because
I
won't
let
you
assign
me
to
it
anyway,.
A
I
wonder
I
mean
remove
label
will
work
just
fine
most
of
the
time.
I
wonder
if
we
have
something
like.
A
A
A
A
A
A
So
I
create
a
sleep
for
I
don't
know
tweet
two
minutes
I'll
disconnect
for
five
minutes
and
then
return.
A
C
D
D
A
Yeah
yeah,
it's
very
likely
that
that
he
just
messed
up
his
cube
cuddle
version
cubeconfig
file.
A
A
The
error
that
he's
getting
is
basically
keep
conflict,
having
issues
with
reading
configuration,
but
at
the
same
time
we
could
theoretically
be
a
little
bit
more
mindful
when
providing
the
information
instead
of
instead
of
giving.
Oh,
we
don't
know,
no
put
kind
is
registered
for
version.
We
want
99
of
time.
We
will
want
to
read
cube
conflicts,
so
the
error
could
be
instead
of
no
kind
could
be
turned
into
no
valid
config
cube,
config
configuration
file
found
or
something
along
those
lines.
D
Yeah,
I
think
I
think,
the
only
time
trying
to
think
about
how
I
would
do
this
if,
if
a
cube
config
is
provided,
then
assume
it's
a
cube,
config
like
a
path
to
a
cube
config
and
then,
but
if
it's,
I
guess
I
would
never
look
at
the
default
location.
So
I
don't
know
I'll
just
look
at
it,
but
you're
right
and
a
warning.
A
better
warning
than
this
would
be
better.
A
But
the
invocation
will
be
in
the
cube
cuddle
side
of
things
where
we
read
the
config,
so
we
could
theoretically
wrap
the
error
and
say
no
valid
cubeconfig
found
and
then,
after
call
on,
we
could
say
no
component.
Is
there
just
infrastructure
if
you
want,
so
that
would
be
like
a
detailed
information
about
what
was
wrong,
but
the
initial
error
would
be.
Oh
you
just
your
cube.
Conflict
is
just
wrong.
Yeah
awesome.
Thank
you.
A
What
he's
asking
is,
instead
of
manually
typing
the
patch
file,
to
be
able
to
specify
a
path
to
a
particular
file,
a
particular
field
and
a
value
why
he
can
just
write
this
thing
on
his
own.
A
A
A
Or
he's
just
asking
about
the
ability
to
specify
to
present
the
the
spec
data
from
the
from
those
fields,
so
we
would
be
showing
I
don't
know.
A
We
would,
for
example,
could
be
showing
the
current
usage
slash.
Well
he's
not
using
slashes,
we
could
use
slash
and
then
you
would
be
showing
requests
or
limits
or
both.
A
I
I'm
not
kind
of
I'll
probably
add
them
on
under
single
cell.
So
if
you
were
asking
cpu,
we
would
show
you
the
current
usage,
the
requests
and
the
limits
something
similar
to
what,
when
we
were
showing
how
many
parts
are
running
in
a
let's
say,
deployment.
We're
showing
that
there's.
For
example,
two
out
of
four,
what
he's
asking
is
to
be
able
to
see?
Oh,
my
pot
is
using
5
out
of
the
10
or
15
resources
that
I
assigned
it
to.
A
Yeah,
we
could
just
not
show
them,
so
we
would
say
15
out
of
something
out
of
or
we
will
be
selling
15
out
of,
I
don't
know
request
limits.
I
would
have
to
check
if
there,
if
something
similar,
we're
already
presenting
somewhere
else,
and
that
seems
like
a
reasonable
addition.
A
I'll
probably
just
wait
for
some
more.
A
Votes
towards
implementing
this
kind
of
functionality
like
we
usually
do
for
offer
these
kind
of
features.
A
So
I'll
let
it
sit
here
for
a
bit.
A
Okay,
let's,
let's
hear
back
from
the
original
author,
if
he's
happy
with
it,
it
looks
like
marcus
mark
already
started.
Looking
at
it.
A
A
A
You
might
be
a
little
bit
confused,
so
first
of
all,
he
was
proposing
that
we
document
the
fact
that
this
warning
this
deprecation
warning
is
expected
because
we
are
in
the
process
of
changing
the,
how
the
version
is
presented
and
the
second
and
the
second
issue
that
he
raised
is
when
you
download
the
first
time
your
cube
cuddle.
You
might
not
necessarily
have
a
kudatis
cluster
running,
which
is
a
rare
case,
but
for
someone
new,
it
will
be
a
thing.
A
So
what
ends
up
happening?
Is
you
run
cube,
cuddle
version,
even
the
short
and
then
you'll
hit
the
30
second
timeout,
after
which
you'll
get
the
information
that
oh,
I
couldn't
find
the
cluster
your
cluster,
because
you
did
not
tell
me
any
so.
I
was
talking
with
katrina
about
it.
A
Last
week,
and
we
both
agreed
that
maybe
we
could
try
to
figure
out
a
way
to
limit
the
timeout
for
that
initial
connection,
so
that
it's,
I
don't
know
a
couple
seconds,
barely
so
that
you
can
have
a
quick
turnaround,
a
quick
information
back
that,
oh,
I
wasn't
able
to
read
the
server.
A
Alternatively,
we
could
change
the
the
docs
to
tell
you
to
use
cubecut
version
short
client,
which
is
which
will
show
you
just
the
client
version
without
reaching
for
the
server
and
explain
that.
Oh,
if
you
don't
know
full
server
version,
first
of
all,
you
need
to
have
one
and
then
just
drop
the
client
dash
client
block.
A
B
Something
whenever
I'm
not
connected
to
a
server,
I
I
get
results
immediately.
I
don't
get
a
30
second
timeout
does
is
what
30
second
timeout
is.
Is
that.
A
I
guess
honestly,
I
I
always
have
some
cluster
running,
whether
that
will
be
our
internal
ci
or
something.
So
I
can't
tell
I
I
remember
talking
with
him
about
it
last
week
and
on
slack
and
then
I
forgot
to
check
it.
So
it's
possible
that
maybe,
since
you
already
have
a
cluster,
it
will
try
to
ping,
it
will
get
an
immediate
information
that
it
can't
it
will
fail,
but
the
when
it
tries
to
reach
the
localhost
8080
it
hangs.
A
I
don't
know
I
that's
something
that
we
will
just
have
to
check.
A
Most
often
than
not,
people
will
have
some
cluster
and
in
most
cases
they
will
want
to
know
the
version
of
both
the
client
and
the
server
together,
rather
than
separately
so
yeah,
probably
a
little
bit
of
dogs
updates.
A
So
if
anyone
is
is
interested
in,
that
could
be
actually
a
very
simple
issue.
I'll
leave
some
comments
afterwards,
and
that
is.
A
B
Just
a
quick
question
so
so
this
is
going
to
go
through
particular
all
the
rules
to
see
you
know
for
the
can
I
it's
and
it's
going
to.
If
you
do
like
a
v10
it'll
tell
you
the
reason
you
know
what
which
particular
rule
allowed
it
or
if
it's
blocked
it'll,
say
none
of
the
rules.
So
so
it's
looking
through
all
of
the
rbac.
A
And
it's
not
it's
not
necessarily
looking
through
alba
our
back.
What
it
does
it
issues.
A
Yes,
one
of
these
self,
I
remember.
A
That's
the
thing
that
I
don't
know
yes,
self.
Subject
rules
review
so
it
issues
a
a
request
over
to
the
server
to
tell
you
do
we
support
it
or
or
is
it
possible
that
I
can't
have
access
or
not
and
in
return
you'll
get
information
about
which
resources
you
have
or
not
have
access,
but
quickly
skimming
through
self-subject
access
review
and
self-subject
rules
review?
B
Require
a
lot
of
work
on
the
server
which
might
be
difficult
to
to
get
through
to.
I
know.
A
What
you
mean,
yes,
is
definitely
something
that
would
be
useful.
I
don't
know
when
or
who
would
want
to
pick
it
up,
but
it's
a
valid
I'll
ask
definitely.
B
A
A
A
So
my
guess
is
that
I'm
not
100
sure
what
kind
of
error
we
are
returning
here,
but
I
would
expect
that
for
an
unknown
time,
we
should
throw
an
error,
as
I'm
not
registered
there,
in
which
case
we
should
try
to
fall
back
to
open
api
information
and
fall
back
to
the
approach
that
we
usually
have
with
crs
any
other
crs
which
are
not
registered
in
the
in
the
cert
on
the
server.
A
So
something
apparently,
I
would
start
with
looking
at
that
error
being
returned
here
for
a
type
that
is
a
built-in
but
a
newer
version.
A
B
Yeah,
I
think
what
is
antoine
saying
there
he's
saying
it:
we
always
use
strategic,
merge
patch
for
the
built-in
types
and
json
merge
for
the
custom
resources.
A
B
So
just
a
quick
note,
antoine
and
some
of
the
folks
in
the
api
machinery
have
been
updating
the
open
api
so
that
there's
actually
an
end
point
for
the
for
v3
now
and
yeah.
There's
there's
been
some
movement
there.
If
you
and
I
can't-
I
don't
know
exactly
what
that
movement
is,
but
there
has
been
some
changes
in
open
api.
B
A
Just
playing
around
or
testing
stuff
because
I
would
say,
use
it
in
production
and
we
were
talking
about
eddie,
had
a
eddie
spent
about
five
minutes
during
the
session
last
week.
Talking
about
the
differences
between
imperative
and
declarative
commands,
and
we
were
pointing
a
lot
of
people
like
everyone
all
in
that
in
that
session
to
use
apply
in
production
imperative
command,
although
the
learning
curve
is
easy,
are
definitely
not
recommended
for
a
production
use.
A
So
that's
yeah,
that's
definitely
something
so,
but
other
than
that.
I
think
what
he's
struggling
with
is,
if
you
use
cue
cutter
like
this,
what
it
does
it
will
actually
set
the
image
on
his
pod
and
print.
The
yam,
which
apply
probably
will
be
the
output
will
contain
information
that
it
got
properly
modified
and
then
the
output
yam.
A
D
A
A
Let's
try
one
another.
We
still
have
about
five
minutes
left
108.584.
A
A
A
A
So
the
output
or
the
information
provided
by
sig
by
q
panel
is
perfectly
correct.
A
Okay,
I'll
I'll
leave
a
comment
there
after
this
call.
A
When
you
specify
a
resource
type
and
then
after
a
space
the
resource
name,
there
are
some
issues
with
howie
pars.
Apparently
the
dash
n
is
being
ignored,
but
the
xyz
is
being
picked
as
the
name
for
the
sds
or
the
dash
n
is
being
picked
or
it
just
ignores
this,
and
then
the
namespace
is
always
pick
the
last
one,
because
you
can't
specify
multiple
namespaces.
It's
always
the
last
one.
So
it's
possible
that
what
happens
in
his
case
is
the
first
namespace
is
being
ignored.
A
The
last
namespace
is
being
picked
so
he's
getting
added,
sts
all
the
sdss
and
the
kafka
namespace.
That's
my
suspicion.
I
could
probably
double
check,
but
yeah.
A
I
see
that
pack
already
commented
that
it's
most
likely
not
about
I'll
double
check
what
I
just
said,
and
something
tells
me
that
we
will
be
able
to
close
this
one
as
well.
A
Okay,
we
got
last
one
minute
and
any
other
closing
comments
that
people
want
to
share.