►
From YouTube: Kubernetes SIG CLI 20230329 - 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
B
A
I'm
gonna
start
off
with
this
one
again,
I
told
this
person
that
this
is
one
that
we
went
over
at
some
point.
I
guess
it's
the
last
bug
scrub.
A
The
answer
was
that
it's
an
imperative
command
and
we
don't
want
to
allow
this
and
they
kept
pushing
it
and
I
said:
okay,
fine
I'll,
we'll
talk
about
it
again,
I'm
pretty
sure
that
the
answer
is
still
no,
but
basically
the
gist
of
it
is
that
they
want
to
be
able
to
allow
deletion
of
resources
specified
by
name
across
all
their
namespaces
I.
B
A
See
any
of
us
changing
our
answers
from
last
time,
but
I
told
them
that
I
would
bring
it
up
so
I'm
bringing
it
up.
A
Something
hold
on
sorry.
I
should
have
read
this
before
I.
Actually,
oh,
that's
right.
They
tried
to
apply
a
label
using
the
cube,
Kettle
label
service
and
then
using
that
and
specifying
that
by
name
and
they
basically
get
the
same
thing
and
I
said
well,
yeah,
that's
the
dashes.
A
Not
this
is
now
how
it's
meant
to
be
consumed.
Basically,
you
need
to
use,
apply
and
use
declarative
commands
and
sorry
yeah
declare
you
the
declarative,
workflows
when
you
get
to
this
point.
A
C
A
Yeah,
that's
pretty
much
how
I
feel
about
it
too,
and,
as
we've
all
said,
multiple
times,
we
don't
really
want
to
extend
out
the
imperative
commands
and
have
people
relying
on
those
instead
of
the
declarative,
workflows
that
have
been
developed
and
should
be
have
documentation
everywhere,
saying
that
that
should
be
the
for
production
workloads.
That's
what
you
should
be
doing.
A
F
E
Is
the
recommended
way
to
delete
things,
because
the
declarative
way
is
the
pruning
thing
that
we're
working
on
improving
right
now,
but
that
has
not
been
in
good
shape
for
a
long
time.
E
But
the
same
argument
applies:
I
think
it's
just
like
they
can
script
it
if
they
want
to,
and
we
don't
want
to
optimize
it's
a
matter
of
like
what
do
we
want
to
optimize
for
here.
Do
we
think
this
is
something
that
everyone
wants
to
do
all
the
time,
and
that
belongs
as
a
feature
in
in
this
core
tool,
and
also
for
deletion.
E
Specifically
what's
proposed
is
something
that
would
have
a
huge
blast
radius
and
we
already
have
lots
of
issues
with
people
unintentionally
having
huge
blast,
radiuses
and
anything
across
namespaces
I
think
there
could
potentially
be
a
performance
argument
for
avoiding
building
it
in.
A
A
So
we
are
going
to
be
working
her
running
through
this
list
here.
Thank
you,
Machi,
and
this
is
gonna,
be
the
first
issue
that
we're
looking
at
one
three,
nine
nine.
A
Okay,
completion
does
not
work
for
create
job
Dash
Dash
from
completion
currently
does
not
work
at
all
after
a
dash
dash
from
flag
immediately
after
the
flag,
it
only
shows
file
names
which
is
incorrect
and
not
resource
types
after
a
resource
type
and
a
slash,
it
doesn't
suggest
any
anything.
I
did
not
read
that
correctly.
A
Okay,
after
a
resource
type
in
a
slash,
it
doesn't
suggest
anything,
let's
say
except
anything
instead
of
the
expected
behavior
of
suggesting
resource
names
why
this
is
needed
create
from
is
used
to
manually,
trigger
cron
jobs
are
not
aware
of
other
use
cases
but
I'm
sure
they
exist
yeah.
So
it
sounds
like
that
looks
like
Mark
is
already
on
on
the
case.
A
F
B
B
A
C
Another
issue
that
was
closed
and
Mark
was
I,
think
it
was
Mark,
was
commenting
on
it
and
recommended
that
they
open
another
issue
for
this
specific
problem.
A
B
A
A
A
B
C
A
C
C
A
Yeah
at
the
bottom
of
us
to
make
sure
that
it's
with
like
a
completion
script,
is
it
in
the
first
entry
in.
A
Still
looks
for
all
plugins
that
would
explain
why
things
remain
slow
when
the
WSL
path
contains
all
the
windows
paths.
Okay,
so
yeah
markedly
seems
to
think
that
this
is
the
problem
and
generally
inclined
to
believe
Mark.
Oh
yeah,
yeah,
okay,
well,
yeah,
I
think
that
we
can
go
ahead
and
just
accept
this.
A
A
Okay,
and
then
here
is
the
next
issue
optimize
the
help
context
when
command
has
a
sub
command
I,
don't
know
how
much
we
want
to
get
into
this,
but
it
is
kind
of
we've
had
issues
with
like
the
website,
not
completing
the
stuff
correctly
for,
for
example,
this
exact
set
of
issues
here,
because
they
weren't
actually
walking
the
path
of
the
command
tree.
A
I
can't
I
meant
to
go
look
at
this
actually
one
because
I
think
this
tagged
me
in
it
somewhere,
but
is
the
usage
for
this
defined
by
like
automatically
defined
by
cobras.
That's
just
something
that
we
have
like
in
there
automatically.
B
Because
I
can't,
let
me
look
actually
exactly.
B
A
Okay,
yeah,
let's
I'm
pretty
sure
it's
not,
let's
see
create
a
secret.
C
C
Cobra
lets
you
override
things
in
ways
that
we're
not
using,
but
I'm,
not
I'm,
not
sure
in
this
specific
case,
if
it
does.
B
A
Yeah
all
right,
I
think
the
the
what
what
the
well
I
guess.
I
actually
don't
know
what
there's
no
keep
pillow
to
have
installed
right
now
hold
on
yeah
I'm
on
125,
so
I'm
a
major
version
behind,
but
I
think
what
the
issue
is
is
that
it's
we're
getting
the
help
text
back,
which
is
fine,
but
the
specific
issue
that
this
person
has
with
it
is
that
the
usage
listed
as
cube
Kettle
create
secret
and.
B
A
Is
a
general
sort
of
secret
not
for
like
secret
generic
or
Secret
TLS
or
whatever,
but
that
it
gives
us
back.
This
Cube
cuddle
create
secret,
and
then
it
says
past
flags
and
then
options
and
the
reality
of
the
situation
is
that
there's
there
are
sub
commands
that
you
can
pass
as
well
as
flags
and
options.
So
it's
more
than
just
what's
listed
in
the
usage
part.
B
C
A
A
A
It
seems
like
a
pretty
easy
one:
I,
don't
know
I
I.
The
thing
is
that
we
might
want
to
check
on
more
places
than
just
that
if
we're
gonna
try
and
do
that
so
I
think
it
might
just
be.
What
it's
doing
is
it's
like
walking
back
up
from
the
command
from
the
secret
command
itself,
and
then
it
goes.
It
finds
the
parent
create
and
the
parent
Cube
Kettle
and
just
like
appends
all
of
those
together.
A
Basically
so
it
could
be
dangerous
and
then
just
adds
flags
and
options
at
the
end,
because
we
couldn't
get
ourselves
stuck
in
a
situation
where
I
mean
I,
don't
see
us
having
sub
commands
under
acute
Kettle
create
secret
like
generic,
for
example,
but
yeah
I,
don't
know,
maybe
at
the
leaves
leaves
of
each
of
the
trees.
We
could
have
something
or
not
the
leaves,
but
the
leaves
minus
one.
E
I
just
tried
adding
the
sub
command
like
the
little
string
sub
command
to
that
place.
You
found
in
the
code
and
that
does
successfully
modify
the
usage
string
as
expected,
but
I
agree
with
your
concern
that,
like
this
probably
isn't
the
only
place
so
you'd
want
some
sort
of
a
audit
of
like
where
that
information
is
missing
and
handle
it
consistently.
A
The
other
thing
is:
do
we
change
it
for
create?
Because
what's
going
to
happen,
does
it
just
merge
all
those
strings
together,
because
then
we
would
have
to
for
the
usage?
Because
then
does
that
become
Cube
Kettle
sub
command,
create
sub
commands,
secret,
sub
command,
I
don't
know,
and
we
can
work.
C
A
B
B
B
A
Basic
cost
was
removed
in
v11.19,
cubicle
config
set
credentials,
command
remains
I,
guess
we're
what
five
seven
seven
major
releases
post
that
so
we
could
go
ahead
and
just
delete
that.
F
F
Script
because
they
they
happen
to
do
that,
so
the
deprecation
is
there
to
announce
that
it
will
be
gone,
so
the
the
flags
or
commands
will
disappear
come
to
help
and
whenever
someone
is
using
the
command,
they
will
get
a
warning
that
this
will
be.
This
will
be
removed
in
a
future
release.
F
So
this
is
this
is
ensuring
that
the
data
base
there
for
sufficient
amounts
of
time,
and
only
then
we
can
drop
it
even
if
it's
doing
nothing,
which
was
the
case
a
couple
of
times
in
the
past,
we
have
to
deprecate
and
follow
the
usual
path.
G
A
This
is
from
Paco,
add
kubernetes
feature
enabled
metric
series
to
track
whether
each
active
feature
guidance
enabled
yes,
I,
remember
getting
a
bunch
of
emails
because
I
got
added
on
this
as
well.
G
G
A
A
So,
if
I'm,
remembering
correctly
who's,
this
Daniel
yeah
Daniel
has
kept
open
for
this
issue.
So.
A
I
think
yeah
okay,
so
Paco
made
PR
to
do
something.
Yeah
so
add
new
command
for
show
feature
Gates
and.
A
It
just
was
the
Daniels
already
working
on
this
and
had
a
cap
open
for
it
and
because
this
was
going
to
require
some.
A
What's
the
word
work
on
the
API
server
side
because
leave
he
said
that
it's
there's
not
really
a
good
way
to
know
this
right
now
and
especially
like
we're
used
they're
using
the
the
metrics.khs.io
pinpoint
and
that's
not
a
good
way
to
figure
out
whether
features
are
enabled,
or
you
know,
that
kind
of
stuff
and
that's
kind
of
where
they
landed
so
I,
I
kind
of
think
we
just
triage
accepts
and
then
let
that
be
a
part
of
an
issue
that
the
Kepp
is
pointing
to.
But
I'm
not
sure
about
that.
E
I
I
don't
know
about
accepting
it
actually
because
the
the
concern
on
this
issue
that
the
data
isn't
actually
accurate
to
what
we
want,
and
this
would
have
to
stay
in
Alpha,
basically
pending
the
totally
different
implementation
in
the
cap,
and
it
sounded
like
there
is
a
way
to
get
the
data.
It's
just
not
very
convenient,
so
I
I,
don't
know
if
it's
worth
implementing
a
temporary
full
feature
that
is
based
on
data
that
we
don't
actually
recommend.
People
rely
on.
A
Okay
yeah
well,
I
was
so
my
I
was
thinking
that
we
would
leave
I.
Guess
it's
like
not
great
to
just
leave
an
industry
open
for
forever,
while
we're
waiting
on
something
else,
though
so
yeah
we
can
just
let
the
cat
kind
of
drive
it
I.
Suppose.
E
E
Specifically
about
the
metric,
so
I
don't
think
the
pr
would
cause
this
issue
to
be
accepted
and
resolved.
It
would
cause.
The
underlying
you
know
feature
request
which
is
is
that
they
want
the
information
about
the
about
the
feature
Gates
to
be
available
in
a
totally
different
way.
If
I
understand
correctly,
do
you
have
any
more
context
on
this?
Perhaps.
F
I
I
only
remember
from
last
week
that
that
Daniel
presented
the
topic
I'm
just
giving
quickly
through
the
enhancements.
I
want
to
know
what
the
API
for
that
will
look
like,
because,
if
it'll
be
just
yet
another
resource
that
we
will
be
reporting
in
a
cluster,
then
just
getting
that
resource
would
be
the
simplest
solution.
F
So
I'm
tempted
to
say
close.
The
issue
and
point
to
the
cap,
because
literally
the
cap
is
solving
this
and.
A
A
F
Other
is
there's
also
a
question
if
the
metric
itself
is
is
an
alpha
metric
or
it's
a
final
one.
Oh
final,
stable,
that's
a
separate
question,
probably
I,
don't
see
an
FPL
apis.
F
Api
section
of
the
cap
says
obviously
we
will
add
an
API
object,
feature
details
below
so
yeah.
You
will
be
able
to
Cube
cuddle,
get
whatever
you
care
about,
so
that
seems
like
probably
the
best
option
for
us.
A
A
A
A
One
solution
would
be
to
make
the
service
code
more
generic
to
support
any
resource
with
a
pod.
Selector
floater
means
the
same.
Look
up
the
resource
check
if
it
has
a
selector,
if
so,
select
the
pods
on
a
custom
resource
when
customer
resources
create
Services
which
the
user
wants
to
or
for
them
they
need
to
look
up
the
name
of
the
service
that
is
owned
by
the
custom
resource.
F
The
problem
is
that
we
currently
have
that
kind
of
information
about
those
thoughts
and
lectures
hard-coded
in
the
incube
cuddle
such
that
Cube
cuddle
looks
at
oh
I'm,
working
with
a
deployment,
so
I
know
that
in
deployment
in
this
pack
in
the
plot
template.
This
is
the
field
that
I
care,
and
this
is
what
I'm
going
to
apply
to
be
able
to
do
something
like
that
with
a
crd.
We
would
have
to
have
to
have
a
way
to
say.
Oh,
this
field
is
a
pot
selector.
F
It
would
have
to
come
up
with
some
kind
of
standardization
about
this
is
a
a
pot
selector
or
something
along
those
lines
to
be
able
to
to
reuse
or
make
the
mechanism
generic,
but
I'm,
not
sure
if
we
want
to.
If
you
want
to
introduce
something
like
that.
F
Theoretically,
we
did
something
like
that
in
the
past,
with
the
scale
sub
resource
to
any
resource
that
and
wants
to
support
the
scaling
it
has
to
implement
the
scales
of
resource
theoretically.
I
could
see
something
like
that
being
done
for
the
Pod
driver,
but
I'm
not
sure
if,
if
it's
worth
the
hassle.
E
F
Yeah
we
can
be
consumers,
but
that
kind
of
standardization
would
have
to
be
done.
Other
elsewhere,
most
likely.
A
Yeah,
okay,
yeah
personally.
For
me
it
sounds
like
they
have
a
solution.
It's
just
like
in
ideal,
but
okay,
so
close.
A
A
Okay,
the
next
issue
is
describe
node
pod
resources,
some
incorrect,
slash,
confusing
yeah.
Probably
you
have
a
node
with
a
pod
which
has
limits
to
unset
in
some
containers.
A
This
is
incorrect,
confusing
in
two
ways
the
sum
over
the
limits
can
be
lower
than
the
requests
in
total.
The
Pod
is
shown
as
Limited
in
its
Global
resource
usage,
while
it
isn't
due
to
an
unset
limit
in
a
container
expectation,
is
a
container
in
pod
does
not
have
a
limit
set
on
a
resource
that
the
Pod
and
total
is
also
not
shown
as
Limited
in
the
usage
of
the
resource
as
the
resource
usage
of
a
pod
as
a
whole
is
also
not
limited.
A
D
D
A
Okay
and
then
Brian
found
the
thing
describe
nodes
with
multi-containers
May
display
error
quota.
A
D
A
C
Yeah-
and
there
was
actually
a
lot
of
comments
on
that
from
from
Daniel
Smith
and
Clayton
Coleman
as
well
so
I
I,
don't
know
it's
been
so
long
or
it's
been
a
while
I
guess
not
so
long,
but
about
maybe
less
than
a
little
less
than
a
year.
C
So
I
guess
we
could
read
those
read
through
those
comments
and
resurrect
this
ER
or
or
whatever.
If
it's,
if
it
is
the
same
issue
but
I
think
I
think
it
might
be
the
same
issue.
A
A
Yeah,
okay,
maybe
do
we
want
to
talk
about
it
on
the
bi-weekly
instead,.
H
Yeah
put
it
on
the
agenda
and
then
here
I'll,
put
on
the
agenda
and
then
it'll
give
us
some
time
to
read
it
before
the
meeting.
C
Yeah
I
can
I'll
have
to
refresh
my
memory.
It's
been
a
while,
but
yeah
yeah,
okay,.
D
A
F
Yeah,
it
was
one
of
the
default
commands,
but
we
had
to
cut
it
out
due
to
it
relying
on
internal
apis.
A
B
H
Probably
a
valid
request
and
it
should
be
simple
to
skip
the
resources
and
the
the
gbrs
that
don't
match
what
they're
trying
to
convert.
A
Like
like
for
the
core
ones
like
service,
if
they
don't
have
the
proper,
what
do
you
call
it
version.
A
C
I
mean:
is
there
a
risk
that
we,
if
they
did,
let's
say
they?
Let's
just
say
they
have
only
one
resource
in
the
file
and
they
try
to
convert
it
and
we
skip
it
and
don't
give
them
an
error
or
anything
they
think
it
worked.
C
H
B
D
F
A
While
deployment
is
paused
after
a
rollout
to
a
new
replica,
set,
started
modifying
the
max
surge
value
or
result
in
incorrect
scaling
of
the
old
replica
set
instead
of
the
new
one,
so
they
create
this
by
reproduce
it
by
creating
a
deployment
edit.
The
deployment
to
change
an
environment
variable
within
the
container
pause
the
deployment
immediately
after
and
then
edit
the
deployment
to
increase
the
max
surge
value.
Watch,
the
old
replica
set
being
scaled
out.
D
A
Yeah
all
right,
okay,
so
I.
A
Assume,
what's
happening
is
because
the
rollout
is
in
the
process
of
happening,
and
you
change
the
surge
for
the
deployment
and
that
gets
applied
to
just
the
deployment
at
large.
That
means
that
the
existing
deployment
will
be
scaled
out
to
fit
the
max
surge
value.
F
This
is
more
slightly
eight,
an
issue
for
Sig
apps
I
think
it's
very
likely
that
the.
F
F
The
max
surge,
on
the
other
hand,
will
be
accounted
as
yet
another
added
operations,
it'll
be
it
should,
if
I
remember
correctly,
be
queued
and
performed
as
the
next
step,
most
likely.
I.
Don't
remember.
On
top
of
my
head
the
deployments,
what
I
would
expect
that
that
would
be
the
case.
What.
F
Can't
remember
I'm
not
sure
if,
because
deployment
has
a
pause
pause
field,
so
it
will
set
that
field
I
think
the
replica
set
has
a
symbol
of
field,
so
it
will
automatically
set.
Both
of
these
I
mean
I
can
quickly
check
it
out.
F
So
so
here's
what
I'm
suspecting
is
happening
since
the
replicasa
does
not
have
anything
by
pause.
He
modifies
a
deployment
by
adding
the
the
environment
created.
Then
he
pauses
the
deployment.
So
the
deployment
controller
will
do
nothing
and
will
not
respond
to
any
changes
on
a
deployment
and
he's
modifying
the
same
deployment
again
adding
the
max
surge,
but
the
replica
set
is
being
scaled
out
as
it
was
before,
because
only
the
deployment
controller
is
passed.
The
yeah
only
the
deployment
is
fast
enough
to
replicas
it.
F
F
C
H
A
A
A
There
are
some
apps
with
checks
whether
a
pod
has
particular
label
present
or
not.
Since
Cube
petal
debug
creates
a
pod
that
doesn't
inherit
labels.
This
is
resulting
in
a
crash
Loop
backup
for
the
actual
container.
That
is
meant
to
be
debugged.
G
I,
don't
think
so.
We
are
clearing
labels
when
debug
comment
is
executed,
but
we
have
plans
to
write
a
cap
for
the
custom,
profiling
and
I
think
we
can
also
handle
this
case
instead
copying
okay
as
an
option
to
set
it
what
they
were
labeled.
They
want
in
this
custom
profiles
and
this
custom
profiling
can
be
used.
Also
for
that
purpose,.
H
G
G
Yes,
but
I'm
not
sure
cat
will
totally
cover
this
or
not.
Maybe
we
can
close
this
one.