►
From YouTube: Kubernetes SIG CLI 20211027 - kustomize 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
Okay,
good
morning,
good
evening,
good
afternoon,
depending
on
where
you
are
today
is
october
27th-
and
this
is
another
of
our
monthly
box
crops
for
customize.
This
time
our
house
will
be
katrina,
although
I'll
be
sharing
my
screen
to
help
her
with
this.
So
take
it
away
from
here.
Katrina.
B
Great
thanks,
michael
before
we
get
started
with
the
triage
board,
does
everyone
anyone
have
an
issue
that
they
wanted
to
highlight
for
us
to
take
a
look
at
in
particular.
B
Not
hearing
anyone,
so
let's
just
dive
in
with
that
first
one,
it
doesn't
actually
look
familiar
and
we
haven't
looked
at
it
in
a
year.
Apparently
so,
let's,
let's
get
this
one
taken
care
of
here,
run
time
for
I'm
pretty.
C
All
right,
I
looked
at
this
like
a
month
ago,
so
the
commit
they're
talking
about
parses,
the
open
api
schema
to
get
name
space
ability
and
I
merged
a
pr
that
basically
undoes
that
commit
but
they're
saying
that
the
performance
is
still
bad.
B
Yeah,
okay,
so
that
column
sort
thing
is,
is
not
doing
what
I
expected.
I
guess.
B
C
B
I
actually
already
replied
to
this
one.
This
morning
I
haven't
had
a
chance
to
work
on
this,
yet
the
problem
is
apparently
the
yaml
v2
library
that
customize
loads,
the
customization
with,
and
so
we
lose
the
quotes
up
front
before
we
even
have
a
chance
to
do
anything
with
them.
So
that's
that's
going
to
be
a
thorny
issue,
but
I
accepted
it
because
I
agree
with
the
author
that,
like
there's
nothing,
they
can
do
to
avoid
this
problem
when
they're
using
customized
but
they're,
then
changing
the
results.
B
There's
possibility
that
the
solution
would
be
in
cube
controller
b
in
the
yellow
library
itself,
but
I
don't
think
it's
something
the
end
user
could
solve
for
themselves.
So
I
accepted
it.
A
Yeah,
isn't
that
something
that
we
already
covered
a
couple
of
times?
I
think
there's
at
least
a
that's
like
third
or
fourth
issue,
where
we're
discussing
the
topic
of
what
is
integers
lose
the
the
quotes
during
the
conversions.
B
Yeah,
I
think
it's
come
up
both
like
this.
This
issue
came
up
in
a
previous
bug
scrub,
but
we
didn't
actually
triage
it,
so
it
came
up
again
well
today,
I
guess
it's.
The
cash
isn't
refreshed
on
the
on
the
board
and
I
think
it's
also
come
up
in
a
cube
kettle
of
bugs
grab
separately.
Like
there's
been
a
separate
issue.
That
is
more
generic
about
where
we,
we
say
to
quote
the
script
strings
manually
as
a
workaround,
but
this
is
saying
well
in
customized.
B
Oh
weird,
this
looks
like
we
actually
that
I
responded
and
probably
in
a
previous
book
scrub.
This
is
about
a
month
ago.
B
B
Yeah
we
already
covered
this
one
in
the
last
bug
scrub
too,
but
we
hadn't
triaged
it
because
jeff
is
taking
care
of
all
the
anchor
issues
and
I
know
jeff
and
I
were
both
very
busy
with
kubecon
preparing
for
keep
going
being
at
cubecon
recovering
from
keep
going.
So
I
imagine
he
has
not
a
chance
to
dig
in
and
confirm
this
one
as
well,
though
it's
very
likely
a
real
issue.
The
anchor
expansion
is
something
released
shortly
before
the
previous
bug
scrub.
It's
a
relatively
new
feature
for
us
well
restored
feature.
D
B
That
you
confirmed
the
the
there
are
two
issues,
but
this
one
about
anchor
tags
that
we
retitled
this
to
match
and
then
there's
another
one
about
integer
keys,
which
is
actually
the
same
issue
that
we
just
you're
confirming
that
the
second
issue,
the
integer
keys,
was
the
same
as
a
different
as
the
other
ticket.
But
I
don't
know
if
anyone
did.
You
also
confirm
the
merge.
B
D
B
A
I
think
he
opened
several
several
of
those
because
he
is
pointing
at
the
anchor
support.
Similarly
to
the
other
issue,.
D
B
B
Yeah
I
wish
jeff
was
here
today
to
ask
him
if
you
just
had
a
chance
to
look
at
any
of
these
follow-up
issues
through
the
anchors,
because
he
was
taking
care
of
this
problem.
It
seems
like
a
bunch
of
the
issues
are
going
to
be
related
to
that.
A
It
would
be
reasonable
to
point
him
to
to
the
plugins
repo
that
you're
setting
up
for
customize
and
just
explain
him
that
he's
open
to
create
those
plugins,
and
he
can
just
add
it
after
he's
done
with
with
it
to
the
index.
I
that
that's
at
least
how
I'm
reading
this.
B
Yeah,
I
guess
another
thing
that
we'll
need
to
point
out
is
that
we're
deprecating
go
plugins,
specifically
we
added
after
our
coupon
presentation
or
before
it
rather
a
bunch
of
new
docs
explaining
which
plug-in
mechanisms
are
deprecated
and
how
to
do
the
various
how
to
build
for
the
various
mechanisms
that
we
intend
to
continue
supporting.
So
that's
something
that
yeah
I'll
assign
this
to
myself
to
respond
in
more
detail.
C
I'm
hesitant
to
add
more
support
for
that
feature.
Just
because
we've
been
talking
about
removing
it.
A
Is
there
a
potential
alternative
approach
that
he
can
take
to
solve
his
problem
that
we
can
point
him
to
instead
of
using
the
helm,
chart
inflation
generator.
C
It
also
looks
like
well:
the
field
helm,
chart,
inflation
generator
is
already
deprecated,
maybe
the
helm,
charts
field
that
replaces
it
supports
this,
but
I'm
not
sure
this
wouldn't
be
that
hard
to
support.
I
think.
B
A
Yeah
and
make
sure
to
explain
that
the
hunter
generator
is
deprecated.
A
Definitely
worth
repeating
if
people
hasn't
haven't
noticed
that
before.
C
B
There
are
also
other
helm
generator
implementations.
The
one
that
we
built
in
is
not
the
only
one.
So
that's
another
thing
we
could,
I
don't
know
natasha
the
the
one
that
kept
built
could
support
this
yeah.
I
can.
C
Point
them
to
that
one!
If,
if
we're
okay
with
that.
B
D
B
I'm
wondering
why
they're
ending
up
with
multiple
config
maps
from
this
in
the
first
place,
because
they
have
the
behavior
create
and
base,
but
then
all
the
other
ones
say
merge.
B
B
Oh,
is
this
because
they're
using
a
name
suffix
in
one
of
the
layers
there,
so
you
end
up
with
a
config.
C
A
B
Yeah,
I
think
I
agree
with
eddie
that
merger
would
not
be
the
solution
that
we
would
recommend.
I
expect
this
will
become
a
support
issue
where
we're
trying
to
figure
out
what
to
recommend
that
they
should
do
with
their
repo
structure.
C
C
She
I
can't
ask
her.
I
think
this
is
a
very
small
change,
so
I
think
she
would
be
fine
with
that.
C
C
Oh,
I
remember
this
one
somewhere
they're
having
like
the
name
space
is
just
empty,
like
it
says,
namespace
colon
and
then.
B
C
A
I
remember
correctly:
we
weaker,
we
will
treat
empty
name
space,
as
is
which
basically
means
as
if
you've
never
unspecified
one.
So
if
you
have
one
particular
namespace
in
your
config,
it
will
inherit
that
one
or
or
whatever
you
have
through
fax.
A
D
B
It
does
sound
like
it,
it's
probably
not
valid,
but
I
agree
with
much
of
that,
like
we
shouldn't
panic
when
we're
given
invalid
data
trying
to
read
through
the
back
trace
here,
see
when
it's
happening,
it
looks
like
potentially
very
girly
or
no
name
reference
transformer.
A
B
Yep,
I
just
threw
the
resources
from
the
test
case
in
and
as
input
and
like
didn't,
do
anything
else,
but
import
them
in
the
file
and
it
blows
up.
So
I'm
going
to
try
to
accept
this
kind
bug
and
yeah
that
is
probably
fairly
approachable.
To
like
I'll
just
comment
that
we
shouldn't
panic.
We
don't
need
to
do
something
great
here,
because
it's
not
valid
the
input
isn't
valid,
but
I
really
shouldn't
panic.
C
A
Katrina
bill
volunteered
for
the
previous
one
42
40.
B
I
wasn't
able
to
assign
you
because
you're
not
an
org
member
yet,
but
I
tagged
you
on
the
issue,
so
hopefully
that
gives
you
a
breadcrumb
to
it.
B
Deliberately
prevent
something
extremely
similar
to
that
right
with
the
loader
restrictions,
where
you
can't
reference
a
customization,
that's
just
somewhere
else
on
your
local
machine.
B
A
B
I'll
respond
to
this
after
I
need
to
look
up
where
in
the
docs
we
explain
that.
A
B
B
I'm
not
sure
if
what
they're
trying
to
do
should
work,
but
the
results
they're
getting
are
strange.
If,
if
that's
in
fact
what's
happening,
it
says
one
of
them
gets
a
properly
suffix
configuration
it's
referenced
and
the
other
one.
It
doesn't
pick
up
the
suffix
at
all,
which
is
that's
inconsistent,
so
that
shouldn't
be
happening.
A
Well,
the
inconsistency
can
be
explained
by
the
fact
that
he
has
conflicts
and
resources
in
one
place,
but
he
doesn't
and
the
other
base
customization
young,
not
sure
not
being
an
expert
in
customized
can,
whether
it's
sufficient
to
have
it
just
at
the
root
and
that's
it,
and
it
should
pick
up
properly
everything
or
whether
he
should
have
this
line.
Also
over
here.
B
A
B
Yeah,
that
is,
unfortunately,
an
error
that
comes
up,
and
that
was
the
original
message
when
we
didn't
support
anchors
at
all
as
well.
It's
not
a
very
helpful
error.
B
B
B
I'm
understanding
correctly
that
they
want.
They
basically
want
to
merge
the
contents
of
those
under
a
single
config
map
key
instead
of
multiple
configmap
keys.
B
Yeah,
it
is
also
possible
to
find
a
key
to
set
a
name
different
from
the
file
name.
I
don't
know
if
that's
going
to
work
if
you
set
the
key
the
same
key
multiple
times,
but
that
is
in
our
documentation.
B
Actually,
the
secret
generator
docs
don't
mention
that
option.
But
if
I
remember
the
code
correctly,
both
are
generated
the
same
way
so
that
the
option
likely
exists
I'll
go
confirm
that
before
I
write
the
comment,
though,.
A
Isn't
the
problem
here
that
we
are
going
through
everything
with
the
name
of
engines,
and
this
is
confusing
that
he's
using
the
engine's
name
for
both
the
conflict
map
and
the
I'll
end
the
deployment
as
well,
although,
theoretically
it
shouldn't
conflict,
it
seems
like
that's.
The
culprit.
B
Yeah,
I
think
the
person
who
commented
below
is
likely
correct
in
that
we
we
use
a
static
set
of
paths
to
figure
out
where
we
should
be
replacing
when
we
do
the
name
transformation
like
where
to
target,
and
we
are
in
fact
targeting
that
path,
even
though,
in
this
example,
it
looks
like
like
the
config
map,
isn't
even
what
that
resource
name
field
is
talking
about.
B
B
I'm
not
sure
how
we
could
handle
that
properly,
given
that
these
rules
target
multiple.
B
B
B
I
think
this
is
incorrect.
Actually,
if
you
look
at
here's
the
pr
that
I
found,
let's
paste
it
in
chat,
it's
called
out
as
being
a
generic
reference,
and
yet
we
accepted
the
patch
that
assumes
it's
a
config
map.
B
C
So
I
guess
like
if
it
if
it
can
be
a
config
map,
then
wouldn't
there
be
cases
where
we
want
it
to
get
the
suffix
like.
How
will
we
differentiate.
B
The
problem
that
I
see
with
this
is
that
it
can
target
so
it
can
target
by
by
name,
but
I
can
also
target
multiple
resource
kinds
of
that
name.
So
say
we
have
a
deployment
and
a
config
map
that
all
in
the
input
have
named
nginx,
and
we
have
this
rule
that
targets
anything
that
deployments
can
figure
out
staple
sets,
etc,
etc.
With
name
engine
x.
B
We
can't
just
transform
that
name
right,
even
though
it
legitimately
does
refer
to
the
config
map
in
the
input,
because
it
also
refers
to
the
deployment
in
the
input
which
now
doesn't
have
the
same
name
like
a
big
map
anymore.
So,
even
if
we
were
to
check
that
field
and
say
like
oh,
yes,
this
is
indeed
targeting
the
v1
group
and
kind
config
map.
B
We
would
also
have
to
verify
that
that
it
targets
that
exclusively
before
doing
a
mutation
and
then,
if
it
doesn't,
then
we
would
have
to
blow
up
like
we
would
have
no
way
of
making
it
correct.
If
I'm
understanding
correctly-
and
I
don't
think
even
that
logic
of
like
let's
look
at
some
other
fields
and
see
what
their
values
are
and
make
a
decision
based
on
this
aggregate
information
about
whether
or
not
to
transform
like-
I
don't
think,
that's
supported
by
the
the
transformer
that
handles
this.
B
I
would
say
it
would
be
better
to
not
be
doing
this
like
it's
more
harmful
to
have
the
transformation,
assuming
it's
always
a
config
map
when
it
very
well
might
not
be,
and
it
very
well
might
be,
targeting
multiple
things
than
it
than
it
is
to
sometimes
not
target.
It
can
pick
up
when
we
should
be
because,
like
we're
supposed
to
be
doing
semantic
transformations
right-
and
in
this
case
it
can't
be
using
the
logic
that
we're
currently
using
to
do
the
replacement.
B
That
would
be
great
thanks,
buddy
all
right,
thanks
everyone
for
coming
today,
we
got
through
a
lot
of
issues
and
see
you
at
the
next
one.